)]}'
{"specs/newton/approved/nova-policies.rst":[{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":13,"context_line":"This spec reconsider (or goes beyond) the `select-destinations-filter`_"},{"line_number":14,"context_line":"blueprint. In the current implementation of watcher, we are bypassing the"},{"line_number":15,"context_line":"rules of the nova scheduler."},{"line_number":16,"context_line":"This may result in failure or even worse, the violation of the nova policies"},{"line_number":17,"context_line":"(affinity rules, capacity, ...)."},{"line_number":18,"context_line":"This is clearly not sustainable, watcher strategies have to be conform to"},{"line_number":19,"context_line":"the administrators specifications."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_70f7dad4","line":16,"range":{"start_line":16,"start_character":63,"end_line":16,"end_character":76},"updated":"2016-06-15 16:22:26.000000000","message":"nova placement policies","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":13,"context_line":"This spec reconsider (or goes beyond) the `select-destinations-filter`_"},{"line_number":14,"context_line":"blueprint. In the current implementation of watcher, we are bypassing the"},{"line_number":15,"context_line":"rules of the nova scheduler."},{"line_number":16,"context_line":"This may result in failure or even worse, the violation of the nova policies"},{"line_number":17,"context_line":"(affinity rules, capacity, ...)."},{"line_number":18,"context_line":"This is clearly not sustainable, watcher strategies have to be conform to"},{"line_number":19,"context_line":"the administrators specifications."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_b110fd5a","line":16,"range":{"start_line":16,"start_character":63,"end_line":16,"end_character":76},"in_reply_to":"7aa08908_70f7dad4","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":11,"context_line":"https://blueprints.launchpad.net/watcher/+spec/nova-policies"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"This spec reconsider (or goes beyond) the `select-destinations-filter`_"},{"line_number":14,"context_line":"blueprint. In the current implementation of watcher, we are bypassing the"},{"line_number":15,"context_line":"rules of the nova scheduler."},{"line_number":16,"context_line":"This may result in failure or even worse, the violation of the nova policies"},{"line_number":17,"context_line":"(affinity rules, capacity, ...)."},{"line_number":18,"context_line":"This is clearly not sustainable, watcher strategies have to be conform to"},{"line_number":19,"context_line":"the administrators specifications."},{"line_number":20,"context_line":""},{"line_number":21,"context_line":"Problem description"},{"line_number":22,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_e5014a06","line":19,"range":{"start_line":14,"start_character":11,"end_line":19,"end_character":34},"updated":"2016-06-15 16:22:26.000000000","message":"I think you should move this as an introduction in the \u0027Problem Description\u0027 section.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":11,"context_line":"https://blueprints.launchpad.net/watcher/+spec/nova-policies"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"This spec reconsider (or goes beyond) the `select-destinations-filter`_"},{"line_number":14,"context_line":"blueprint. In the current implementation of watcher, we are bypassing the"},{"line_number":15,"context_line":"rules of the nova scheduler."},{"line_number":16,"context_line":"This may result in failure or even worse, the violation of the nova policies"},{"line_number":17,"context_line":"(affinity rules, capacity, ...)."},{"line_number":18,"context_line":"This is clearly not sustainable, watcher strategies have to be conform to"},{"line_number":19,"context_line":"the administrators specifications."},{"line_number":20,"context_line":""},{"line_number":21,"context_line":"Problem description"},{"line_number":22,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_71169559","line":19,"range":{"start_line":14,"start_character":11,"end_line":19,"end_character":34},"in_reply_to":"7aa08908_e5014a06","updated":"2016-06-16 09:51:24.000000000","message":"IHMO, the aim of this part is to propose in a single paragraph the aim of this spec.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":21,"context_line":"Problem description"},{"line_number":22,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":23,"context_line":""},{"line_number":24,"context_line":"The Filter Scheduler uses the \u0027select_destinations\u0027 method to decide where to"},{"line_number":25,"context_line":"create or move an instance across the cloud compute infrastructure. It is used"},{"line_number":26,"context_line":"for placement of a new instance as well as migration of an existing instance."},{"line_number":27,"context_line":"In the `select-destinations-filter`_ blueprint,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_9c8ee7ce","line":24,"range":{"start_line":24,"start_character":31,"end_line":24,"end_character":50},"updated":"2016-06-15 16:22:26.000000000","message":"Jed, could you add the link to the source code https://github.com/openstack/nova/blob/3316fa85668ee96f4db1bbff0752de25ce4744af/nova/scheduler/driver.py#L61","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":21,"context_line":"Problem description"},{"line_number":22,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":23,"context_line":""},{"line_number":24,"context_line":"The Filter Scheduler uses the \u0027select_destinations\u0027 method to decide where to"},{"line_number":25,"context_line":"create or move an instance across the cloud compute infrastructure. It is used"},{"line_number":26,"context_line":"for placement of a new instance as well as migration of an existing instance."},{"line_number":27,"context_line":"In the `select-destinations-filter`_ blueprint,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_f10a8526","line":24,"range":{"start_line":24,"start_character":31,"end_line":24,"end_character":50},"in_reply_to":"7aa08908_9c8ee7ce","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":30,"context_line":"this solution leads to several problems."},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_cb97ffe9","line":33,"range":{"start_line":33,"start_character":4,"end_line":33,"end_character":25},"updated":"2016-06-15 16:22:26.000000000","message":"``select_destinations``","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":30,"context_line":"this solution leads to several problems."},{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_6fc9895a","line":33,"range":{"start_line":33,"start_character":4,"end_line":33,"end_character":25},"in_reply_to":"7aa08908_cb97ffe9","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_50d8de7f","line":35,"range":{"start_line":35,"start_character":4,"end_line":35,"end_character":15},"updated":"2016-06-15 16:22:26.000000000","message":"https://github.com/openstack/nova/blob/faeede29795d1eeee241d50f8521431e30ed540f/nova/objects/request_spec.py#L33","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_8fbe75ac","line":35,"range":{"start_line":35,"start_character":4,"end_line":35,"end_character":15},"in_reply_to":"7aa08908_50d8de7f","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_45b2de07","line":36,"range":{"start_line":34,"start_character":0,"end_line":36,"end_character":10},"updated":"2016-06-15 16:22:26.000000000","message":"To migrate an instance, we have to invoke the \u0027select_destinations\u0027 method by passing, as parameter, the associated instance RequestSpec object.\n\n..code:: python\n\n    @abc.abstractmethod\n    def select_destinations(self, context, spec_obj):\n        \"\"\"Must override select_destinations method.\n        :return: A list of dicts with \u0027host\u0027, \u0027nodename\u0027 and \u0027limits\u0027 as keys\n            that satisfies the request_spec and filter_properties.\n        \"\"\"\n        return []","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Before diving into the problems, let\u0027s make clear the behaviour of"},{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ecc2aa85","line":36,"range":{"start_line":34,"start_character":0,"end_line":36,"end_character":10},"in_reply_to":"7aa08908_45b2de07","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"},{"line_number":40,"context_line":"policies, the requested destination is returned, otherwise the NoValidHost"},{"line_number":41,"context_line":"exception is raised. Nova throws this error also when it fails to find the"},{"line_number":42,"context_line":"appropriate target host for an instance"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_50461e22","line":39,"range":{"start_line":36,"start_character":12,"end_line":39,"end_character":25},"updated":"2016-06-15 16:22:26.000000000","message":"Before to call this method, we can now set an unique destination host by updating the RequestSpec object parameter named requested_destination (new in object version 1.6)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":33,"context_line":"the \u0027select_destinations\u0027 method."},{"line_number":34,"context_line":"In order to invoke the \u0027select_destinations\u0027 method, we have to first retrieve"},{"line_number":35,"context_line":"the RequestSpec object `request-spec-object`_ of the instance that we want"},{"line_number":36,"context_line":"to migrate. Then, we can specify in the RequestSpec the requested destination."},{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"},{"line_number":40,"context_line":"policies, the requested destination is returned, otherwise the NoValidHost"},{"line_number":41,"context_line":"exception is raised. Nova throws this error also when it fails to find the"},{"line_number":42,"context_line":"appropriate target host for an instance"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_6cb87ae9","line":39,"range":{"start_line":36,"start_character":12,"end_line":39,"end_character":25},"in_reply_to":"7aa08908_50461e22","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"},{"line_number":40,"context_line":"policies, the requested destination is returned, otherwise the NoValidHost"},{"line_number":41,"context_line":"exception is raised. Nova throws this error also when it fails to find the"},{"line_number":42,"context_line":"appropriate target host for an instance"},{"line_number":43,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_3024325f","line":40,"range":{"start_line":40,"start_character":0,"end_line":40,"end_character":8},"updated":"2016-06-15 16:22:26.000000000","message":"placement policies","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":37,"context_line":"We can only specify one Destination at a time nova/scheduler/host_manager.py."},{"line_number":38,"context_line":"Finally, we can invoke the \u0027select_destinations\u0027 method with the modified"},{"line_number":39,"context_line":"RequestSpec as parameter. If the requested destination satisfies the nova"},{"line_number":40,"context_line":"policies, the requested destination is returned, otherwise the NoValidHost"},{"line_number":41,"context_line":"exception is raised. Nova throws this error also when it fails to find the"},{"line_number":42,"context_line":"appropriate target host for an instance"},{"line_number":43,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_4c081eda","line":40,"range":{"start_line":40,"start_character":0,"end_line":40,"end_character":8},"in_reply_to":"7aa08908_3024325f","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":42,"context_line":"appropriate target host for an instance"},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"The first problem is that the \u0027select_destinations\u0027 method is an internal API"},{"line_number":45,"context_line":"and It is well known that this is not a good practice to invoke RPC API"},{"line_number":46,"context_line":"directly, especially between two components that do not belong to the same"},{"line_number":47,"context_line":"Openstack project. However, Jay Pipes has recently added the blueprint"},{"line_number":48,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_10e016fa","line":45,"range":{"start_line":45,"start_character":4,"end_line":45,"end_character":6},"updated":"2016-06-15 16:22:26.000000000","message":"s/It/it","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":42,"context_line":"appropriate target host for an instance"},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"The first problem is that the \u0027select_destinations\u0027 method is an internal API"},{"line_number":45,"context_line":"and It is well known that this is not a good practice to invoke RPC API"},{"line_number":46,"context_line":"directly, especially between two components that do not belong to the same"},{"line_number":47,"context_line":"Openstack project. However, Jay Pipes has recently added the blueprint"},{"line_number":48,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ccc9ae7b","line":45,"range":{"start_line":45,"start_character":4,"end_line":45,"end_character":6},"in_reply_to":"7aa08908_10e016fa","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":48,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"},{"line_number":49,"context_line":"\u0027select_destinations\u0027 as a RESTFul API."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"In the next sentences, we are assuming that we can call \u0027select_destinations\u0027"},{"line_number":52,"context_line":"method from Watcher through this RESTFul API."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_30a9d291","line":51,"range":{"start_line":51,"start_character":12,"end_line":51,"end_character":21},"updated":"2016-06-15 16:22:26.000000000","message":"s/sentences/part","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":48,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"},{"line_number":49,"context_line":"\u0027select_destinations\u0027 as a RESTFul API."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"In the next sentences, we are assuming that we can call \u0027select_destinations\u0027"},{"line_number":52,"context_line":"method from Watcher through this RESTFul API."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_8cb7a6f4","line":51,"range":{"start_line":51,"start_character":12,"end_line":51,"end_character":21},"in_reply_to":"7aa08908_30a9d291","updated":"2016-06-16 09:51:24.000000000","message":"I like sentences :)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":52,"context_line":"method from Watcher through this RESTFul API."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"},{"line_number":55,"context_line":"\u0027select_destinations\u0027 method has validated a host, the nova scheduler is"},{"line_number":56,"context_line":"virtually consuming the resources of the instance on the selected host until"},{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_10b3f6ea","line":56,"range":{"start_line":55,"start_character":70,"end_line":56,"end_character":19},"updated":"2016-06-15 16:22:26.000000000","message":"vitually consumes (or reserves)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":52,"context_line":"method from Watcher through this RESTFul API."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"},{"line_number":55,"context_line":"\u0027select_destinations\u0027 method has validated a host, the nova scheduler is"},{"line_number":56,"context_line":"virtually consuming the resources of the instance on the selected host until"},{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_7148558b","line":56,"range":{"start_line":55,"start_character":70,"end_line":56,"end_character":19},"in_reply_to":"7aa08908_10b3f6ea","updated":"2016-06-16 09:51:24.000000000","message":"this is the vocabulary by nova.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"},{"line_number":55,"context_line":"\u0027select_destinations\u0027 method has validated a host, the nova scheduler is"},{"line_number":56,"context_line":"virtually consuming the resources of the instance on the selected host until"},{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"},{"line_number":60,"context_line":"the Nova Scheduler will return the selected host and then,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_504f3ebe","line":57,"range":{"start_line":57,"start_character":44,"end_line":57,"end_character":78},"updated":"2016-06-15 16:22:26.000000000","message":"Could you add a HTTP link: https://github.com/openstack/nova/blob/cfd64c976b8aaacaf09fd2a6303877674fcfada0/nova/scheduler/host_manager.py#L182","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":54,"context_line":"The second problem is the tracking of the resources. In fact, each time the"},{"line_number":55,"context_line":"\u0027select_destinations\u0027 method has validated a host, the nova scheduler is"},{"line_number":56,"context_line":"virtually consuming the resources of the instance on the selected host until"},{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"},{"line_number":60,"context_line":"the Nova Scheduler will return the selected host and then,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ec3d8a2d","line":57,"range":{"start_line":57,"start_character":44,"end_line":57,"end_character":78},"in_reply_to":"7aa08908_504f3ebe","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"},{"line_number":60,"context_line":"the Nova Scheduler will return the selected host and then,"},{"line_number":61,"context_line":"it will try to start the instance on it, as soon as possible."},{"line_number":62,"context_line":"In Watcher this is not the case. The strategy developer can create a large"},{"line_number":63,"context_line":"scope of VM placement algorithms thanks to his pluggable architecture"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_6b708b74","line":60,"range":{"start_line":60,"start_character":31,"end_line":60,"end_character":34},"updated":"2016-06-15 16:22:26.000000000","message":"an unique","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":57,"context_line":"the information is updated per compute node(nova/scheduler/host_manager.py:182)"},{"line_number":58,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":59,"context_line":"workflow. In fact, each time the \u0027select_destinations\u0027 method is called,"},{"line_number":60,"context_line":"the Nova Scheduler will return the selected host and then,"},{"line_number":61,"context_line":"it will try to start the instance on it, as soon as possible."},{"line_number":62,"context_line":"In Watcher this is not the case. The strategy developer can create a large"},{"line_number":63,"context_line":"scope of VM placement algorithms thanks to his pluggable architecture"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_2ce4b2aa","line":60,"range":{"start_line":60,"start_character":31,"end_line":60,"end_character":34},"in_reply_to":"7aa08908_6b708b74","updated":"2016-06-16 09:51:24.000000000","message":"It depends of the scheduler implementation. Sometime, we can several hosts.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":70,"context_line":"virtual machine migrations in order to reach the audit goals (minimize energy"},{"line_number":71,"context_line":"consumption, load balacing, etc)."},{"line_number":72,"context_line":"Therefore, in a Watcher strategy it makes sense to evaluate a migration action"},{"line_number":73,"context_line":"but never  execute this action. So, if we invoke the \u0027select_destinations\u0027"},{"line_number":74,"context_line":"method in the Watcher strategies, the nova scheduler will quickly"},{"line_number":75,"context_line":"(depending of the numbers of available hosts) raise the NoValidHost exception"},{"line_number":76,"context_line":"and may refuse some migration even if it makes sense to temporarily violate"},{"line_number":77,"context_line":"some constraint in order to achieve a given optimization goal."},{"line_number":78,"context_line":""},{"line_number":79,"context_line":"We performed some benchmark in order to evaluate if the Nova-Scheduler"},{"line_number":80,"context_line":"can handle a high number of requests from the Watcher strategies."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_4b202fb1","line":77,"range":{"start_line":73,"start_character":32,"end_line":77,"end_character":62},"updated":"2016-06-15 16:22:26.000000000","message":"So, if we invoke the ``select_destinations``\nmethod from Watcher strategies, the nova scheduler will reserve resources on destination compute node for nothing, if this last is validated. And Watcher will receive quickly a NoValidHost exception from the Nova scheduler. In this case, the strategy will not be able to compute other possible migration.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":78,"context_line":""},{"line_number":79,"context_line":"We performed some benchmark in order to evaluate if the Nova-Scheduler"},{"line_number":80,"context_line":"can handle a high number of requests from the Watcher strategies."},{"line_number":81,"context_line":"The results show that It takes in average 49 ms to validate 10"},{"line_number":82,"context_line":"destinations and the time consumed by the select_destinations to reply"},{"line_number":83,"context_line":"is linearly related to the number of destinations."},{"line_number":84,"context_line":"So, this is clearly not sustainable."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_6b66cbef","line":81,"range":{"start_line":81,"start_character":22,"end_line":81,"end_character":24},"updated":"2016-06-15 16:22:26.000000000","message":"s/It/it","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":78,"context_line":""},{"line_number":79,"context_line":"We performed some benchmark in order to evaluate if the Nova-Scheduler"},{"line_number":80,"context_line":"can handle a high number of requests from the Watcher strategies."},{"line_number":81,"context_line":"The results show that It takes in average 49 ms to validate 10"},{"line_number":82,"context_line":"destinations and the time consumed by the select_destinations to reply"},{"line_number":83,"context_line":"is linearly related to the number of destinations."},{"line_number":84,"context_line":"So, this is clearly not sustainable."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ec874ab4","line":81,"range":{"start_line":81,"start_character":22,"end_line":81,"end_character":24},"in_reply_to":"7aa08908_6b66cbef","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":90,"context_line":""},{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_965d1e18","line":93,"range":{"start_line":93,"start_character":50,"end_line":93,"end_character":60},"updated":"2016-06-15 16:22:26.000000000","message":"ceilometer","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":90,"context_line":""},{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_d67686a2","line":93,"range":{"start_line":93,"start_character":33,"end_line":93,"end_character":45},"updated":"2016-06-15 16:22:26.000000000","message":"nova-compute","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":90,"context_line":""},{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ccab6e1d","line":93,"range":{"start_line":93,"start_character":50,"end_line":93,"end_character":60},"in_reply_to":"7aa08908_965d1e18","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":90,"context_line":""},{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_0ca61653","line":93,"range":{"start_line":93,"start_character":33,"end_line":93,"end_character":45},"in_reply_to":"7aa08908_d67686a2","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_763d122c","line":94,"range":{"start_line":94,"start_character":21,"end_line":94,"end_character":23},"updated":"2016-06-15 16:22:26.000000000","message":"s/is/was","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":91,"context_line":"We performed the experiments on two dedicated Ubuntu 14.04 server with 24"},{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ef5e1978","line":94,"range":{"start_line":94,"start_character":21,"end_line":94,"end_character":23},"in_reply_to":"7aa08908_763d122c","updated":"2016-06-16 09:51:24.000000000","message":"We can use present.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"},{"line_number":98,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_7624f20d","line":95,"range":{"start_line":95,"start_character":21,"end_line":95,"end_character":23},"updated":"2016-06-15 16:22:26.000000000","message":"s/is/was","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":92,"context_line":"cores, 125 GB RAM and a solid state disk. All of the OpenStack services were"},{"line_number":93,"context_line":"running on one server except the Nova-compute and Ceilometer."},{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"},{"line_number":98,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_af4d31c6","line":95,"range":{"start_line":95,"start_character":21,"end_line":95,"end_character":23},"in_reply_to":"7aa08908_7624f20d","updated":"2016-06-16 09:51:24.000000000","message":"ditto","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"},{"line_number":98,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"},{"line_number":99,"context_line":"instance and then see if the Nova-scheduler agrees."},{"line_number":100,"context_line":"If this instance has an affinity constraint, we can try many combinations and"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_d6b846ca","line":97,"range":{"start_line":97,"start_character":0,"end_line":97,"end_character":18},"updated":"2016-06-15 16:22:26.000000000","message":"What is the third ? :)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":94,"context_line":"The scheduler driver is configured with the `CachingScheduler`_ and the"},{"line_number":95,"context_line":"\u0027select_destination\u0027 is exposed through a RESTFul API."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"The fourth problem is that Watcher is blind (it has no way to integrate the"},{"line_number":98,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"},{"line_number":99,"context_line":"instance and then see if the Nova-scheduler agrees."},{"line_number":100,"context_line":"If this instance has an affinity constraint, we can try many combinations and"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_26568e06","line":97,"range":{"start_line":97,"start_character":0,"end_line":97,"end_character":18},"in_reply_to":"7aa08908_d6b846ca","updated":"2016-06-16 09:51:24.000000000","message":"héhé :) I was hungry","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"The last problem is that we can\u0027t use the of \u0027select_destinations\u0027 to validate"},{"line_number":106,"context_line":"instances with an affinity group."},{"line_number":107,"context_line":"It\u0027s seem that the soft-affinity-for-server-group blueprint tackle this"},{"line_number":108,"context_line":"problem. We should take a look to"},{"line_number":109,"context_line":"Do we need to see an affinity group has a monolithic instance ?"},{"line_number":110,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_d1069084","line":107,"range":{"start_line":107,"start_character":19,"end_line":107,"end_character":58},"updated":"2016-06-15 16:22:26.000000000","message":"please add the HTTP link: https://blueprints.launchpad.net/nova/+spec/soft-affinity-for-server-group","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"The last problem is that we can\u0027t use the of \u0027select_destinations\u0027 to validate"},{"line_number":106,"context_line":"instances with an affinity group."},{"line_number":107,"context_line":"It\u0027s seem that the soft-affinity-for-server-group blueprint tackle this"},{"line_number":108,"context_line":"problem. We should take a look to"},{"line_number":109,"context_line":"Do we need to see an affinity group has a monolithic instance ?"},{"line_number":110,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_b1f02493","line":107,"range":{"start_line":107,"start_character":60,"end_line":107,"end_character":66},"updated":"2016-06-15 16:22:26.000000000","message":"s/tackle/tackles","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"The last problem is that we can\u0027t use the of \u0027select_destinations\u0027 to validate"},{"line_number":106,"context_line":"instances with an affinity group."},{"line_number":107,"context_line":"It\u0027s seem that the soft-affinity-for-server-group blueprint tackle this"},{"line_number":108,"context_line":"problem. We should take a look to"},{"line_number":109,"context_line":"Do we need to see an affinity group has a monolithic instance ?"},{"line_number":110,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_26f20e07","line":107,"range":{"start_line":107,"start_character":60,"end_line":107,"end_character":66},"in_reply_to":"7aa08908_b1f02493","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":104,"context_line":""},{"line_number":105,"context_line":"The last problem is that we can\u0027t use the of \u0027select_destinations\u0027 to validate"},{"line_number":106,"context_line":"instances with an affinity group."},{"line_number":107,"context_line":"It\u0027s seem that the soft-affinity-for-server-group blueprint tackle this"},{"line_number":108,"context_line":"problem. We should take a look to"},{"line_number":109,"context_line":"Do we need to see an affinity group has a monolithic instance ?"},{"line_number":110,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_c6d18a58","line":107,"range":{"start_line":107,"start_character":19,"end_line":107,"end_character":58},"in_reply_to":"7aa08908_d1069084","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":105,"context_line":"The last problem is that we can\u0027t use the of \u0027select_destinations\u0027 to validate"},{"line_number":106,"context_line":"instances with an affinity group."},{"line_number":107,"context_line":"It\u0027s seem that the soft-affinity-for-server-group blueprint tackle this"},{"line_number":108,"context_line":"problem. We should take a look to"},{"line_number":109,"context_line":"Do we need to see an affinity group has a monolithic instance ?"},{"line_number":110,"context_line":""},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_e353ffe0","line":108,"range":{"start_line":108,"start_character":26,"end_line":108,"end_character":33},"updated":"2016-06-15 16:22:26.000000000","message":"s/look to/look at this one.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":113,"context_line":"Use Cases"},{"line_number":114,"context_line":"----------"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"As a strategy developer, I want to have the Nova policies of the instances"},{"line_number":117,"context_line":"(scheduler hints) to make sure that the destination host that"},{"line_number":118,"context_line":"I\u0027m providing are conform to the administrators specifications."},{"line_number":119,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_c3d6c3f0","line":116,"updated":"2016-06-15 16:22:26.000000000","message":"placement","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":114,"context_line":"----------"},{"line_number":115,"context_line":""},{"line_number":116,"context_line":"As a strategy developer, I want to have the Nova policies of the instances"},{"line_number":117,"context_line":"(scheduler hints) to make sure that the destination host that"},{"line_number":118,"context_line":"I\u0027m providing are conform to the administrators specifications."},{"line_number":119,"context_line":""},{"line_number":120,"context_line":"As an administrator trying to optimize his infrastructure, I would like to use"},{"line_number":121,"context_line":"the Nova Scheduler to validate/check the Watcher Audit before applying"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_c32a4390","line":118,"range":{"start_line":117,"start_character":57,"end_line":118,"end_character":62},"updated":"2016-06-15 16:22:26.000000000","message":"defined by my algorithm will be conform with the instance placement policies set by the admin.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":117,"context_line":"(scheduler hints) to make sure that the destination host that"},{"line_number":118,"context_line":"I\u0027m providing are conform to the administrators specifications."},{"line_number":119,"context_line":""},{"line_number":120,"context_line":"As an administrator trying to optimize his infrastructure, I would like to use"},{"line_number":121,"context_line":"the Nova Scheduler to validate/check the Watcher Audit before applying"},{"line_number":122,"context_line":"it to the system."},{"line_number":123,"context_line":""},{"line_number":124,"context_line":""},{"line_number":125,"context_line":"Project Priority"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_8646891f","line":122,"range":{"start_line":120,"start_character":59,"end_line":122,"end_character":17},"updated":"2016-06-15 16:22:26.000000000","message":"I would like to be sure that migration actions proposed by a strategy have been validated on Nova side, before applying them on the cluster.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":19214,"name":"Vincent Mahe","email":"vincent.mahe@b-com.com","username":"vmahe35"},"change_message_id":"e9a438072ad2059dab5703026735633e8c360474","unresolved":false,"context_lines":[{"line_number":129,"context_line":"Proposed change"},{"line_number":130,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"This specification suggests to retrieve a snapshot of the instances policies."},{"line_number":133,"context_line":"In this way, it will be possible to have holistic approach to determine the"},{"line_number":134,"context_line":"optimal cloud datacenter configuration, while meeting the tenants SLAs."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":".. image:: ../../../doc/source/images/nova_watcher.svg"},{"line_number":137,"context_line":"   :width: 140%"},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"This specification suggests to retrieve a snapshot of the instances policies."},{"line_number":140,"context_line":"In this way, it will be possible to have holistic approach to determine the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_3c7a5378","line":137,"range":{"start_line":132,"start_character":0,"end_line":137,"end_character":15},"updated":"2016-06-15 12:27:26.000000000","message":"Should remove this duplicate block of text","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"cb2cf1f0c5cbbb5cae2fef85d37dfe543a9e422c","unresolved":false,"context_lines":[{"line_number":129,"context_line":"Proposed change"},{"line_number":130,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"This specification suggests to retrieve a snapshot of the instances policies."},{"line_number":133,"context_line":"In this way, it will be possible to have holistic approach to determine the"},{"line_number":134,"context_line":"optimal cloud datacenter configuration, while meeting the tenants SLAs."},{"line_number":135,"context_line":""},{"line_number":136,"context_line":".. image:: ../../../doc/source/images/nova_watcher.svg"},{"line_number":137,"context_line":"   :width: 140%"},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"This specification suggests to retrieve a snapshot of the instances policies."},{"line_number":140,"context_line":"In this way, it will be possible to have holistic approach to determine the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_fc8c4b30","line":137,"range":{"start_line":132,"start_character":0,"end_line":137,"end_character":15},"in_reply_to":"7aa08908_3c7a5378","updated":"2016-06-15 12:34:13.000000000","message":"ah yes :)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":136,"context_line":".. image:: ../../../doc/source/images/nova_watcher.svg"},{"line_number":137,"context_line":"   :width: 140%"},{"line_number":138,"context_line":""},{"line_number":139,"context_line":"This specification suggests to retrieve a snapshot of the instances policies."},{"line_number":140,"context_line":"In this way, it will be possible to have holistic approach to determine the"},{"line_number":141,"context_line":"optimal cloud datacenter configuration, while meeting the tenants SLAs."},{"line_number":142,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_c61531e6","line":139,"range":{"start_line":139,"start_character":68,"end_line":139,"end_character":76},"updated":"2016-06-15 16:22:26.000000000","message":"placement policies","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":153,"context_line":"migration action to Nova for validation. It would be much better to send the"},{"line_number":154,"context_line":"ActionPlan compared to only validate the action atomically."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_5cb2b407","line":156,"range":{"start_line":156,"start_character":4,"end_line":156,"end_character":10},"updated":"2016-06-15 16:22:26.000000000","message":"s/cloud/cluster","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":153,"context_line":"migration action to Nova for validation. It would be much better to send the"},{"line_number":154,"context_line":"ActionPlan compared to only validate the action atomically."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_46c8fa75","line":156,"range":{"start_line":156,"start_character":4,"end_line":156,"end_character":10},"in_reply_to":"7aa08908_5cb2b407","updated":"2016-06-16 09:51:24.000000000","message":"I think we can both.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":154,"context_line":"ActionPlan compared to only validate the action atomically."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_5c80f4f7","line":157,"range":{"start_line":157,"start_character":26,"end_line":157,"end_character":34},"updated":"2016-06-15 16:22:26.000000000","message":"validity (or pertinence)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":154,"context_line":"ActionPlan compared to only validate the action atomically."},{"line_number":155,"context_line":""},{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_8699427c","line":157,"range":{"start_line":157,"start_character":26,"end_line":157,"end_character":34},"in_reply_to":"7aa08908_5c80f4f7","updated":"2016-06-16 09:51:24.000000000","message":"I prefer validity (more the notion of time)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"},{"line_number":161,"context_line":"because they have reached their maximum capacity meanwhile."},{"line_number":162,"context_line":"This is the reason why, we will let the Nova-Scheduler confirm that the"},{"line_number":163,"context_line":"computed actions in the Audit are still complying with the Nova constraints."},{"line_number":164,"context_line":"If later on, the audit is stale, we can launch a new audit. Watcher wants to"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_3c38883d","line":161,"range":{"start_line":159,"start_character":0,"end_line":161,"end_character":59},"updated":"2016-06-15 16:22:26.000000000","message":"This part should be in the \"Problem description\". No ?","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":156,"context_line":"The cloud configuration may change over time in the same way as the workloads"},{"line_number":157,"context_line":"of the instances. So, the validity of the proposed actions are subject to a"},{"line_number":158,"context_line":"race condition."},{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"},{"line_number":161,"context_line":"because they have reached their maximum capacity meanwhile."},{"line_number":162,"context_line":"This is the reason why, we will let the Nova-Scheduler confirm that the"},{"line_number":163,"context_line":"computed actions in the Audit are still complying with the Nova constraints."},{"line_number":164,"context_line":"If later on, the audit is stale, we can launch a new audit. Watcher wants to"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_66f7f68c","line":161,"range":{"start_line":159,"start_character":0,"end_line":161,"end_character":59},"in_reply_to":"7aa08908_3c38883d","updated":"2016-06-16 09:51:24.000000000","message":"I will think about it.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"},{"line_number":161,"context_line":"because they have reached their maximum capacity meanwhile."},{"line_number":162,"context_line":"This is the reason why, we will let the Nova-Scheduler confirm that the"},{"line_number":163,"context_line":"computed actions in the Audit are still complying with the Nova constraints."},{"line_number":164,"context_line":"If later on, the audit is stale, we can launch a new audit. Watcher wants to"},{"line_number":165,"context_line":"incrementally improve the system when there is a possible time window,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_7c3a1036","line":162,"range":{"start_line":162,"start_character":40,"end_line":162,"end_character":54},"updated":"2016-06-15 16:22:26.000000000","message":"nova-scheduler service","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":159,"context_line":"As a result, the validated actions can be invalidated 1s later."},{"line_number":160,"context_line":"For example, the selected computes could no longer be accepting new instances"},{"line_number":161,"context_line":"because they have reached their maximum capacity meanwhile."},{"line_number":162,"context_line":"This is the reason why, we will let the Nova-Scheduler confirm that the"},{"line_number":163,"context_line":"computed actions in the Audit are still complying with the Nova constraints."},{"line_number":164,"context_line":"If later on, the audit is stale, we can launch a new audit. Watcher wants to"},{"line_number":165,"context_line":"incrementally improve the system when there is a possible time window,"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_260b2e7f","line":162,"range":{"start_line":162,"start_character":40,"end_line":162,"end_character":54},"in_reply_to":"7aa08908_7c3a1036","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":170,"context_line":"Applier can execute a sub-graph which is capable to react to the error as"},{"line_number":171,"context_line":"gracefully as possible."},{"line_number":172,"context_line":""},{"line_number":173,"context_line":"We could for example use the Nova-Scheduler to pro-actively select a"},{"line_number":174,"context_line":"pre-computed alternative path. However, solving this problem is not in"},{"line_number":175,"context_line":"the scope of this spec."},{"line_number":176,"context_line":"In the same way, we will also need the `automatic-triggering-audit`_"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_bce0f897","line":173,"range":{"start_line":173,"start_character":29,"end_line":173,"end_character":43},"updated":"2016-06-15 16:22:26.000000000","message":"ditto","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":170,"context_line":"Applier can execute a sub-graph which is capable to react to the error as"},{"line_number":171,"context_line":"gracefully as possible."},{"line_number":172,"context_line":""},{"line_number":173,"context_line":"We could for example use the Nova-Scheduler to pro-actively select a"},{"line_number":174,"context_line":"pre-computed alternative path. However, solving this problem is not in"},{"line_number":175,"context_line":"the scope of this spec."},{"line_number":176,"context_line":"In the same way, we will also need the `automatic-triggering-audit`_"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_a648becd","line":173,"range":{"start_line":173,"start_character":29,"end_line":173,"end_character":43},"in_reply_to":"7aa08908_bce0f897","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":171,"context_line":"gracefully as possible."},{"line_number":172,"context_line":""},{"line_number":173,"context_line":"We could for example use the Nova-Scheduler to pro-actively select a"},{"line_number":174,"context_line":"pre-computed alternative path. However, solving this problem is not in"},{"line_number":175,"context_line":"the scope of this spec."},{"line_number":176,"context_line":"In the same way, we will also need the `automatic-triggering-audit`_"},{"line_number":177,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":178,"context_line":"(Watcher will automatically apply the Audit)."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_9caedc2d","line":175,"range":{"start_line":174,"start_character":31,"end_line":175,"end_character":23},"updated":"2016-06-15 16:22:26.000000000","message":"so what will be default behavior if an action can not be successfully completed? We ignore it, we stop the action plan ?","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":171,"context_line":"gracefully as possible."},{"line_number":172,"context_line":""},{"line_number":173,"context_line":"We could for example use the Nova-Scheduler to pro-actively select a"},{"line_number":174,"context_line":"pre-computed alternative path. However, solving this problem is not in"},{"line_number":175,"context_line":"the scope of this spec."},{"line_number":176,"context_line":"In the same way, we will also need the `automatic-triggering-audit`_"},{"line_number":177,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":178,"context_line":"(Watcher will automatically apply the Audit)."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_86df82d5","line":175,"range":{"start_line":174,"start_character":31,"end_line":175,"end_character":23},"in_reply_to":"7aa08908_9caedc2d","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":174,"context_line":"pre-computed alternative path. However, solving this problem is not in"},{"line_number":175,"context_line":"the scope of this spec."},{"line_number":176,"context_line":"In the same way, we will also need the `automatic-triggering-audit`_"},{"line_number":177,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":178,"context_line":"(Watcher will automatically apply the Audit)."},{"line_number":179,"context_line":"It will be also be usefull to have the `specify-multiple-live-migrate-hosts`_"},{"line_number":180,"context_line":"blueprint implemented to directly propose to Nova-Scheduler"},{"line_number":181,"context_line":"several options."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_f701f905","line":178,"range":{"start_line":177,"start_character":48,"end_line":178,"end_character":44},"updated":"2016-06-15 16:22:26.000000000","message":"the making of an action plan, and its immediate execution.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":177,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":178,"context_line":"(Watcher will automatically apply the Audit)."},{"line_number":179,"context_line":"It will be also be usefull to have the `specify-multiple-live-migrate-hosts`_"},{"line_number":180,"context_line":"blueprint implemented to directly propose to Nova-Scheduler"},{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_172485b0","line":180,"range":{"start_line":180,"start_character":45,"end_line":180,"end_character":59},"updated":"2016-06-15 16:22:26.000000000","message":"ditto","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":177,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":178,"context_line":"(Watcher will automatically apply the Audit)."},{"line_number":179,"context_line":"It will be also be usefull to have the `specify-multiple-live-migrate-hosts`_"},{"line_number":180,"context_line":"blueprint implemented to directly propose to Nova-Scheduler"},{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_06a1d246","line":180,"range":{"start_line":180,"start_character":45,"end_line":180,"end_character":59},"in_reply_to":"7aa08908_172485b0","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":180,"context_line":"blueprint implemented to directly propose to Nova-Scheduler"},{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_d28e8b78","line":184,"range":{"start_line":183,"start_character":0,"end_line":184,"end_character":23},"updated":"2016-06-15 16:22:26.000000000","message":"When the watcher-applier service is ready to execute an action plan, it will first check globally if each migration action are validated by the nova scheduler. If an action is no more valid, the action plan will be set to FAILED, and none action will be triggered.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_372381a9","line":184,"range":{"start_line":184,"start_character":43,"end_line":184,"end_character":50},"updated":"2016-06-15 16:22:26.000000000","message":"``execute``","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_a6927ef4","line":184,"range":{"start_line":184,"start_character":43,"end_line":184,"end_character":50},"in_reply_to":"7aa08908_372381a9","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":180,"context_line":"blueprint implemented to directly propose to Nova-Scheduler"},{"line_number":181,"context_line":"several options."},{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_a6a9de26","line":184,"range":{"start_line":183,"start_character":0,"end_line":184,"end_character":23},"in_reply_to":"7aa08908_d28e8b78","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"},{"line_number":188,"context_line":"method."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_52cf3b29","line":185,"updated":"2016-06-15 16:22:26.000000000","message":"please, add a return carriage. + \nBefore Action execution, we should check again the validity of this one ...","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":182,"context_line":""},{"line_number":183,"context_line":"The workflow engine code in watcher/applier/workflow_engine/default.py"},{"line_number":184,"context_line":"will need to be changed. Specifically, the execute method will be amended"},{"line_number":185,"context_line":"to pro-actively check the validity of the Audit. The code of the migration"},{"line_number":186,"context_line":"action (watcher/applier/actions/migration.py#precondition) will be amended to"},{"line_number":187,"context_line":"validate the migration of the instance by calling the select_destinations"},{"line_number":188,"context_line":"method."}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ec5e2778","line":185,"in_reply_to":"7aa08908_52cf3b29","updated":"2016-06-16 09:51:24.000000000","message":"Done","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":188,"context_line":"method."},{"line_number":189,"context_line":""},{"line_number":190,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":191,"context_line":"will be amended to add the get_policies method. We should adapt all existing"},{"line_number":192,"context_line":"strategies so that they take into account the Nova constraints."},{"line_number":193,"context_line":""},{"line_number":194,"context_line":"We shoud provide a helper for the strategies developer to query the nova"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_52245b5f","line":191,"range":{"start_line":191,"start_character":27,"end_line":191,"end_character":39},"updated":"2016-06-15 16:22:26.000000000","message":"You should add the resource manager name (nova, and maybe later cinder, neutron ...) so get_nova_policies or get_policies(\u0027nova\u0027)","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":188,"context_line":"method."},{"line_number":189,"context_line":""},{"line_number":190,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":191,"context_line":"will be amended to add the get_policies method. We should adapt all existing"},{"line_number":192,"context_line":"strategies so that they take into account the Nova constraints."},{"line_number":193,"context_line":""},{"line_number":194,"context_line":"We shoud provide a helper for the strategies developer to query the nova"}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_8f8af562","line":191,"range":{"start_line":191,"start_character":27,"end_line":191,"end_character":39},"in_reply_to":"7aa08908_52245b5f","updated":"2016-06-16 09:51:24.000000000","message":"I have to first look at more precisely how to structure the information. This is not addressed yet in this spec.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":12394,"name":"David TARDIVEL","email":"David.TARDIVEL@b-com.com","username":"David.T"},"change_message_id":"a700b7c1e9ca1d7603ee65ac3efbf26d2e28f5b8","unresolved":false,"context_lines":[{"line_number":200,"context_line":"This spec will also need  that the RequestSpec object to be exposed as"},{"line_number":201,"context_line":"a RESTFul API."},{"line_number":202,"context_line":"We could be also nice to be able to subscribe"},{"line_number":203,"context_line":"Nova / Novaclient / openstackclient may also have to be modified."},{"line_number":204,"context_line":""},{"line_number":205,"context_line":"Do we need to provide our own driver of Nova-Scheduler ?"},{"line_number":206,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_c801bab1","line":203,"updated":"2016-06-15 16:22:26.000000000","message":"we should be able to call select_destinations in \u0027trial\u0027 mode, i.e without resources reservation.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"11507fe3c92caf6463cd7e7ef3c558b44e0c71b3","unresolved":false,"context_lines":[{"line_number":200,"context_line":"This spec will also need  that the RequestSpec object to be exposed as"},{"line_number":201,"context_line":"a RESTFul API."},{"line_number":202,"context_line":"We could be also nice to be able to subscribe"},{"line_number":203,"context_line":"Nova / Novaclient / openstackclient may also have to be modified."},{"line_number":204,"context_line":""},{"line_number":205,"context_line":"Do we need to provide our own driver of Nova-Scheduler ?"},{"line_number":206,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aa08908_ef2f39a5","line":203,"in_reply_to":"7aa08908_c801bab1","updated":"2016-06-16 09:51:24.000000000","message":"I have to complete the spec for this point.","commit_id":"1abeb07070ff1a803783d2bad6d82a5f4a233d13"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_2a1d64be","line":35,"updated":"2016-06-17 14:03:39.000000000","message":"I think there is another case:\n\n - Do we want Watcher to select instances to be moved, and also a preferred set of hosts to move it to, and let Nova Scheduler pick a host from that group that meets any constraints?","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"e0a976364aaf541f3ec0140e2df325d6cb2319dc","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_e103000c","line":35,"in_reply_to":"5a9d85d2_3b9ff904","updated":"2016-06-20 07:52:58.000000000","message":"Ed Leafe:\nYour second point says that Watcher does not select the target host at all.\nI was not enough clear, this is was the purpose of this sentence \u003d\u003e \"We can also propose destinations to nova but without knowing the constraints\"\n\nJoe Cropper:\nI totally agree, This is exactly the point of this specification. I\u0027m also worry when there is many constraints.\n\n\nIMHO, the first approach can be to retrieve the constraints but it seems that we have to wait a long time before that nova provides to us the RESTFul API that we need. So, I\u0027m looking to a new approach, we could defined our own scheduler driver in nova.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"79e73eeb4bbc0429a63ba7ee002b8c62570e37a6","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_f9e4350e","line":35,"in_reply_to":"5a9d85d2_3b9ff904","updated":"2016-06-20 20:00:23.000000000","message":"If you have 10 possible hosts and only 1 is acceptable, with the multiple host change for live migration (https://review.openstack.org/276840), you\u0027d pass in 10, and let the scheduler pick from that. No errors.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"79e73eeb4bbc0429a63ba7ee002b8c62570e37a6","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_f9cef558","line":35,"in_reply_to":"5a9d85d2_e103000c","updated":"2016-06-20 20:00:23.000000000","message":"Watcher knows the Data Center constraints: CPU load; network load, temperatures, etc., that determine if the VMs are optimally arranged. It then makes suggestions for moving VMs in response to the data it knows about.\n\nThe scheduler knows nothing of the things that Watcher knows about. All it knows is when it is asked to move a VM to a host, whether that host can accept that VM.\n\nYou\u0027re trying to make Watcher know everything by essentially duplicating the scheduler\u0027s logic in Watcher. That duplication, as well as trying to hook into undocumented interfaces is what I think is wrong.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"c85d944663f2d651e5493ad688ac6252c8b25ca6","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_e47c77ed","line":35,"in_reply_to":"5a9d85d2_f9e4350e","updated":"2016-06-21 08:01:34.000000000","message":"I agree that your blueprint can be the first step for Watcher. I will highlight this solution in the spec.\n\nNo errors. \u003d\u003e this is very depending of the strategy, the number of constraints and the cluster configuration.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_ac5b3185","line":35,"in_reply_to":"7aa08908_2a1d64be","updated":"2016-06-17 15:18:41.000000000","message":"Okay, this is my second point, but I will replace by your.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"8c7d95a15b39b6157f8dc9056ca6fc9069075842","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_e009b832","line":35,"in_reply_to":"7aa08908_ac5b3185","updated":"2016-06-17 15:35:57.000000000","message":"Your second point says that Watcher does not select the target host at all. I\u0027m saying that Watcher *does* select the target host(s), and all that the scheduler does is verify that one of the targets is acceptable.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"703f7b81cb76cfb85b1cf68cf156a7e66585f3d4","unresolved":false,"context_lines":[{"line_number":32,"context_line":"    destinations to nova but without knowing the constraints)"},{"line_number":33,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":34,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":35,"context_line":"    used ?"},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Problem description"},{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_3b9ff904","line":35,"in_reply_to":"7aa08908_e009b832","updated":"2016-06-20 06:37:08.000000000","message":"The problem I have with this approach is that we don\u0027t realize the host isn\u0027t valid until we *try* the migration and it throws a NVH exception.  If there are 10 hosts, but only 1 valid target... we could get 9 errors before we ever find the one valid option.\n\nDon\u0027t we think there\u0027s a better option for this?  I\u0027m especially worried about when you have a large number of hosts in a data center...\n\nBUT, I suppose we could do this as a starting point, but we just need to tolerate the NVH exceptions when we\u0027re trying the migrations.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Watcher is optimizing the whole infrastructure by looking at the opportunities"},{"line_number":41,"context_line":"(holistic approach). Watcher don\u0027t wants to check every potential placements"},{"line_number":42,"context_line":"with the nova scheduler before making a choice."},{"line_number":43,"context_line":"This approach requires a huge number of queries and it may never be able to"},{"line_number":44,"context_line":"find a viable solution."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_2a7444fc","line":41,"range":{"start_line":41,"start_character":29,"end_line":41,"end_character":40},"updated":"2016-06-17 14:03:39.000000000","message":"doesn\u0027t want","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":38,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"Watcher is optimizing the whole infrastructure by looking at the opportunities"},{"line_number":41,"context_line":"(holistic approach). Watcher don\u0027t wants to check every potential placements"},{"line_number":42,"context_line":"with the nova scheduler before making a choice."},{"line_number":43,"context_line":"This approach requires a huge number of queries and it may never be able to"},{"line_number":44,"context_line":"find a viable solution."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_cccddd31","line":41,"range":{"start_line":41,"start_character":29,"end_line":41,"end_character":40},"in_reply_to":"7aa08908_2a7444fc","updated":"2016-06-17 15:18:41.000000000","message":"thanks","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":73,"context_line":"directly, especially between two components that do not belong to the same"},{"line_number":74,"context_line":"Openstack project. However, Jay Pipes has recently added the blueprint"},{"line_number":75,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"},{"line_number":76,"context_line":"`select_destinations`_ as a RESTFul API."},{"line_number":77,"context_line":""},{"line_number":78,"context_line":"In the next sentences, we are assuming that we can call `select_destinations`_"},{"line_number":79,"context_line":"method from Watcher through this RESTFul API."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_aa403460","line":76,"updated":"2016-06-17 14:03:39.000000000","message":"This is indeed an eventual goal, but the timeframe for getting this as a *public* RESTful API that Watcher could consume is at least a year away, and most likely longer. So all of the following parts of this spec are at best only possible in the distant future. They cannot be part of Watcher now.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":73,"context_line":"directly, especially between two components that do not belong to the same"},{"line_number":74,"context_line":"Openstack project. However, Jay Pipes has recently added the blueprint"},{"line_number":75,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"},{"line_number":76,"context_line":"`select_destinations`_ as a RESTFul API."},{"line_number":77,"context_line":""},{"line_number":78,"context_line":"In the next sentences, we are assuming that we can call `select_destinations`_"},{"line_number":79,"context_line":"method from Watcher through this RESTFul API."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_cf378ffc","line":76,"in_reply_to":"7aa08908_aa403460","updated":"2016-06-17 15:18:41.000000000","message":"We can bet that one day this will happen and to avoid that watcher stay stuck, we could create a plugin that can call the nova-scheduler via RPC.\nWe will also help the nova community to make it happen.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"8c7d95a15b39b6157f8dc9056ca6fc9069075842","unresolved":false,"context_lines":[{"line_number":73,"context_line":"directly, especially between two components that do not belong to the same"},{"line_number":74,"context_line":"Openstack project. However, Jay Pipes has recently added the blueprint"},{"line_number":75,"context_line":"`scheduler-rest-api-select-destinations`_ which aims at exposing the"},{"line_number":76,"context_line":"`select_destinations`_ as a RESTFul API."},{"line_number":77,"context_line":""},{"line_number":78,"context_line":"In the next sentences, we are assuming that we can call `select_destinations`_"},{"line_number":79,"context_line":"method from Watcher through this RESTFul API."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_3baad571","line":76,"in_reply_to":"7aa08908_cf378ffc","updated":"2016-06-17 15:35:57.000000000","message":"I would rather see Watcher designed to work well with Nova as it exists, and then, when it changes sufficiently to warrant investigating a new design, consider the actual new interface when it exists.\n\nThere was a previous effort to separate the scheduler a few years ago (Gantt). It too promised a separate scheduler service with a RESTful API. After a few years of not getting very far, it was dropped. If projects had designed themselves to work with Gantt, they would have wasted their time. I don\u0027t want to see Watcher make that mistake.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":85,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":86,"context_line":"workflow. In fact, each time the `select_destinations`_ method is called,"},{"line_number":87,"context_line":"the Nova Scheduler will return the selected host and then,"},{"line_number":88,"context_line":"it will try to start the instance on it, as soon as possible."},{"line_number":89,"context_line":"In Watcher this is not the case. The strategy developer can create a large"},{"line_number":90,"context_line":"scope of VM placement algorithms thanks to his pluggable architecture"},{"line_number":91,"context_line":"(they can for example use exact approaches and Heuristics or Meta-heuristics)."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_ec109950","line":88,"updated":"2016-06-17 14:03:39.000000000","message":"No, the scheduler doesn\u0027t start an instance. The select_destinations method simply selects a destination. The calling service, Nova conductor, then takes the returned (host, node) tuple and attempts to start the requested VM on that node.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":85,"context_line":"This is not a problem for Nova because this is only one step in a wider"},{"line_number":86,"context_line":"workflow. In fact, each time the `select_destinations`_ method is called,"},{"line_number":87,"context_line":"the Nova Scheduler will return the selected host and then,"},{"line_number":88,"context_line":"it will try to start the instance on it, as soon as possible."},{"line_number":89,"context_line":"In Watcher this is not the case. The strategy developer can create a large"},{"line_number":90,"context_line":"scope of VM placement algorithms thanks to his pluggable architecture"},{"line_number":91,"context_line":"(they can for example use exact approaches and Heuristics or Meta-heuristics)."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_4c1d6dbd","line":88,"in_reply_to":"7aa08908_ec109950","updated":"2016-06-17 15:18:41.000000000","message":"Yes, you are right. I will fix that. \nI wanted mainly to explain that it can\u0027t happen separately.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":101,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":102,"context_line":"resources the nova scheduler will quickly (depending of the numbers of"},{"line_number":103,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":104,"context_line":"the proposed action will be denied even if it possible."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_6ca029f5","line":104,"updated":"2016-06-17 14:03:39.000000000","message":"The scheduler doesn\u0027t consume resources; it simply evaluates hosts based on its current view of the host\u0027s state.\n\nThe issue is that while Watcher is planning its moves, the cloud is still operating, and VMs can be created, moved, and destroyed between the original call to select_destinations, and the eventual call to migrate a VM. This is a classic race condition.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"8c7d95a15b39b6157f8dc9056ca6fc9069075842","unresolved":false,"context_lines":[{"line_number":101,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":102,"context_line":"resources the nova scheduler will quickly (depending of the numbers of"},{"line_number":103,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":104,"context_line":"the proposed action will be denied even if it possible."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_5b87e160","line":104,"in_reply_to":"7aa08908_157d281c","updated":"2016-06-17 15:35:57.000000000","message":"It \"consumes\" only in the local HostManager; it doesn\u0027t consume in the actual hosts. The only \"consumption\" in the scheduler is in the case where it has been asked to find more than one host (for booting multiple instances). That\u0027s why it does a _get_all_host_states at the top of the loop, which queries the Nova DB for the current host states. Inside of the loop, it \"consumes\" from its local copy so that the next pass through the loop doesn\u0027t try to allocate the same resources. It doesn\u0027t touch the host, or the host resource records in the Nova DB.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"e0a976364aaf541f3ec0140e2df325d6cb2319dc","unresolved":false,"context_lines":[{"line_number":101,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":102,"context_line":"resources the nova scheduler will quickly (depending of the numbers of"},{"line_number":103,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":104,"context_line":"the proposed action will be denied even if it possible."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_01ba8ce1","line":104,"in_reply_to":"7aa08908_5b87e160","updated":"2016-06-20 07:52:58.000000000","message":"This is the reason why I wrote that the scheduler is *virtually* consuming the resources. \n\nThe point here, is to say that, if the strategy is calling massively the select_destinations method before the Nova DB is called, nova will raise the NoValidHost exception. Even, if the host hasn\u0027t instances. \n\nIf you want I can share with you the code snippet that highlight the problem.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":101,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":102,"context_line":"resources the nova scheduler will quickly (depending of the numbers of"},{"line_number":103,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":104,"context_line":"the proposed action will be denied even if it possible."},{"line_number":105,"context_line":""},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_157d281c","line":104,"in_reply_to":"7aa08908_6ca029f5","updated":"2016-06-17 15:18:41.000000000","message":"If you take a look at this line:\nhttps://github.com/openstack/nova/blob/master/nova/scheduler/filter_scheduler.py#L136\nand https://github.com/openstack/nova/blob/master/nova/scheduler/filter_scheduler.py#L98\n\nIt says that the scheduler \"consume\" the resource. \n\nHere an concrete example of the problem that I want to explain:\n\nT\u003d Time X \nCluster state \u003d cluster configuration.\n\nT-0 \u003d\u003e Cluster state:\nHost_1 (available 1 vcpu)\nHost_2 (available 4 vcpu)\n\nT-1 \u003d\u003e watcher strategy call select_destinations(request_spec\u003d4vcpu) \n---\u003e selected_host\u003d HOST_2\n\nT-1 \u003d\u003e Cluster state:\nHost_1 (available 1 vcpu)\nHost_2 (available 0 vcpu)\n\nT-2 \u003d\u003e watcher strategy call select_destinations(request_spec\u003d1vcpu) \n---\u003e selected_host\u003d HOST_1\n\nT-2 \u003d\u003e Cluster state:\nHost_1 (available 0 vcpu)\nHost_2 (available 0 vcpu)\n\nT-3 \u003d\u003e watcher strategy call select_destinations(request_spec\u003d1vcpu) \nraise NoValidHost\n\nIf we use select_destinations method in the watcher strategies that doesn\u0027t mean necessarily that we want to do this action, we just want to check if it is possible.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":108,"context_line":"can handle a high number of requests from the Watcher strategies."},{"line_number":109,"context_line":"The results show that it takes in average 49 ms to validate 10"},{"line_number":110,"context_line":"destinations and the time consumed by the `select_destinations`_ to reply"},{"line_number":111,"context_line":"is linearly related to the number of destinations."},{"line_number":112,"context_line":"So, this is clearly not sustainable."},{"line_number":113,"context_line":"You can see below a chart:"},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_accb31a3","line":111,"updated":"2016-06-17 14:03:39.000000000","message":"This is highly dependent on the class of scheduler being used, and the overall activity of the rest of the cloud. Additionally, the presence and number of cells being used will greatly affect performance. Many large deployments shard their cloud using cells.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":108,"context_line":"can handle a high number of requests from the Watcher strategies."},{"line_number":109,"context_line":"The results show that it takes in average 49 ms to validate 10"},{"line_number":110,"context_line":"destinations and the time consumed by the `select_destinations`_ to reply"},{"line_number":111,"context_line":"is linearly related to the number of destinations."},{"line_number":112,"context_line":"So, this is clearly not sustainable."},{"line_number":113,"context_line":"You can see below a chart:"},{"line_number":114,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_e048982f","line":111,"in_reply_to":"7aa08908_accb31a3","updated":"2016-06-17 15:18:41.000000000","message":"I agree, but we should start somewhere. I just want to show that the default scheduler performances doesn\u0027t allow us to use it in the watcher strategy.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":124,"context_line":""},{"line_number":125,"context_line":"The third problem is that Watcher is blind (it has no way to integrate the"},{"line_number":126,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"},{"line_number":127,"context_line":"instance and then see if the Nova-scheduler agrees."},{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_ec48992f","line":127,"updated":"2016-06-17 14:03:39.000000000","message":"Yes!","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"8c7d95a15b39b6157f8dc9056ca6fc9069075842","unresolved":false,"context_lines":[{"line_number":124,"context_line":""},{"line_number":125,"context_line":"The third problem is that Watcher is blind (it has no way to integrate the"},{"line_number":126,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"},{"line_number":127,"context_line":"instance and then see if the Nova-scheduler agrees."},{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_7b373db0","line":127,"in_reply_to":"7aa08908_a00df090","updated":"2016-06-17 15:35:57.000000000","message":"You say \"blind\"; I say \"decoupled\". :)\n\nThe design should be independent services that respect service boundaries.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":124,"context_line":""},{"line_number":125,"context_line":"The third problem is that Watcher is blind (it has no way to integrate the"},{"line_number":126,"context_line":"Nova constraints in its strategies). We can only try a new placement for an"},{"line_number":127,"context_line":"instance and then see if the Nova-scheduler agrees."},{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_a00df090","line":127,"in_reply_to":"7aa08908_ec48992f","updated":"2016-06-17 15:18:41.000000000","message":"Do we want that \"watcher\" to stay blind ?","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"},{"line_number":131,"context_line":"of the nova placement policies and try the combinations internally."},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"The last problem is that we can\u0027t use the of `select_destinations`_ to validate"},{"line_number":134,"context_line":"instances with an affinity group."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_ac5971ff","line":131,"updated":"2016-06-17 14:03:39.000000000","message":"It\u0027s also much less likely to ever be possible to do that.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"8c7d95a15b39b6157f8dc9056ca6fc9069075842","unresolved":false,"context_lines":[{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"},{"line_number":131,"context_line":"of the nova placement policies and try the combinations internally."},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"The last problem is that we can\u0027t use the of `select_destinations`_ to validate"},{"line_number":134,"context_line":"instances with an affinity group."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_dbe4710e","line":131,"in_reply_to":"7aa08908_0021644d","updated":"2016-06-17 15:35:57.000000000","message":"Several reasons:\n\n1) having the scheduler expose its configured filtering strategy will probably never happen.\n\n2) being able to read both the request spec, flavor table, and resource tables will require a major change to the Nova API, and that\u0027s one that I know many (including myself) would not be in favor of.\n\n3) Race conditions. While Watcher is trying to mimic the scheduler, the cloud is changing, and now instead of just a single racy process (scheduler), you now have two racy processes.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":128,"context_line":"If this instance has an affinity constraint, we can try many combinations and"},{"line_number":129,"context_line":"never find a viable solution."},{"line_number":130,"context_line":"It would be much more efficient for the Watcher strategies to fetch a snapshot"},{"line_number":131,"context_line":"of the nova placement policies and try the combinations internally."},{"line_number":132,"context_line":""},{"line_number":133,"context_line":"The last problem is that we can\u0027t use the of `select_destinations`_ to validate"},{"line_number":134,"context_line":"instances with an affinity group."}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_0021644d","line":131,"in_reply_to":"7aa08908_ac5971ff","updated":"2016-06-17 15:18:41.000000000","message":"Why ?","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"703f7b81cb76cfb85b1cf68cf156a7e66585f3d4","unresolved":false,"context_lines":[{"line_number":211,"context_line":"After this step if the Audit is not stale we can be confident for his"},{"line_number":212,"context_line":"application."},{"line_number":213,"context_line":"However, depending of the number of Actions that contains the"},{"line_number":214,"context_line":"Audit his application can take time."},{"line_number":215,"context_line":"So, we will verify again atomically the Action. In this way, the code of the"},{"line_number":216,"context_line":"Action `migration`_ will be amended to call the `select_destinations`_ method"},{"line_number":217,"context_line":"just prior the execution."}],"source_content_type":"text/x-rst","patch_set":4,"id":"5a9d85d2_bb4e49bf","line":214,"range":{"start_line":214,"start_character":6,"end_line":214,"end_character":9},"updated":"2016-06-20 06:37:08.000000000","message":"s/his/this","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"0b2e99a02c5edd4a31b8c96785ea190427b6b547","unresolved":false,"context_lines":[{"line_number":322,"context_line":"Elements that should be discussed with the Nova team:"},{"line_number":323,"context_line":"WIP"},{"line_number":324,"context_line":"This spec will need that the `select_destinations`_ method to be exposed as a"},{"line_number":325,"context_line":"RESTFul API."},{"line_number":326,"context_line":""},{"line_number":327,"context_line":"Nova / Novaclient / openstackclient may also have to be modified."},{"line_number":328,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_0ccfe57f","line":325,"updated":"2016-06-17 14:03:39.000000000","message":"Reiterating the point made above: this *will not* happen for at least a year, and most likely longer. It may never happen.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"245f319d22e91ed44aac2afb3747e44e1552a072","unresolved":false,"context_lines":[{"line_number":322,"context_line":"Elements that should be discussed with the Nova team:"},{"line_number":323,"context_line":"WIP"},{"line_number":324,"context_line":"This spec will need that the `select_destinations`_ method to be exposed as a"},{"line_number":325,"context_line":"RESTFul API."},{"line_number":326,"context_line":""},{"line_number":327,"context_line":"Nova / Novaclient / openstackclient may also have to be modified."},{"line_number":328,"context_line":""}],"source_content_type":"text/x-rst","patch_set":4,"id":"7aa08908_60d82854","line":325,"in_reply_to":"7aa08908_0ccfe57f","updated":"2016-06-17 15:18:41.000000000","message":"see my previous my comment.","commit_id":"6fea02aec4818306c8e190cee34a7181692270a9"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"dde9ae3d2e915dbee256612ab39c053f9b9dcbdc","unresolved":false,"context_lines":[{"line_number":45,"context_line":"with the Nova Scheduler before making a choice. This approach requires a huge"},{"line_number":46,"context_line":"number of queries and it may never be able to find a viable solution."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The Filter Scheduler uses the `select_destinations`_ method to decide where to"},{"line_number":49,"context_line":"create or move an instance across the cloud compute infrastructure. It is used"},{"line_number":50,"context_line":"for placement of a new instance as well as migration of an existing instance."},{"line_number":51,"context_line":"In the `select-destinations-filter`_ blueprint,"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_ab72420f","line":48,"range":{"start_line":48,"start_character":4,"end_line":48,"end_character":10},"updated":"2016-06-28 13:55:57.000000000","message":"nit: not only the FilterScheduler driver, but rather all of them, including the ChanceScheduler...","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"5545122c28ac0a6c1a45d4ea7b55fc2fed4a976b","unresolved":false,"context_lines":[{"line_number":45,"context_line":"with the Nova Scheduler before making a choice. This approach requires a huge"},{"line_number":46,"context_line":"number of queries and it may never be able to find a viable solution."},{"line_number":47,"context_line":""},{"line_number":48,"context_line":"The Filter Scheduler uses the `select_destinations`_ method to decide where to"},{"line_number":49,"context_line":"create or move an instance across the cloud compute infrastructure. It is used"},{"line_number":50,"context_line":"for placement of a new instance as well as migration of an existing instance."},{"line_number":51,"context_line":"In the `select-destinations-filter`_ blueprint,"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_c1768d3d","line":48,"range":{"start_line":48,"start_character":4,"end_line":48,"end_character":10},"in_reply_to":"3aaa91ec_ab72420f","updated":"2016-06-29 06:45:03.000000000","message":"yes, you are right. I will fix that.","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"dde9ae3d2e915dbee256612ab39c053f9b9dcbdc","unresolved":false,"context_lines":[{"line_number":78,"context_line":"`select_destinations`_ as a RESTFul API."},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"In the next sentences, we are assuming that we can call `select_destinations`_"},{"line_number":81,"context_line":"method from Watcher through this RESTFul API."},{"line_number":82,"context_line":""},{"line_number":83,"context_line":"The second problem is the tracking of the resources. In fact, each time the"},{"line_number":84,"context_line":"`select_destinations`_ method has validated a host, the Nova Scheduler is"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_cbe50ea9","line":81,"updated":"2016-06-28 13:55:57.000000000","message":"TBC, it\u0027s not yet a Newton bp, even probably not an Ocata too...","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"dde9ae3d2e915dbee256612ab39c053f9b9dcbdc","unresolved":false,"context_lines":[{"line_number":105,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":106,"context_line":"resources the Nova Scheduler will quickly (depending of the numbers of"},{"line_number":107,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":108,"context_line":"the proposed action will be denied even if possible."},{"line_number":109,"context_line":""},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_2ef660f9","line":108,"updated":"2016-06-28 13:55:57.000000000","message":"Right, that\u0027s exactly why using select_destinations() is not good, because the nova community explains that it\u0027s not for a dryrun.\n\nInstead, you should rather ask all the resources by the REST API and try to figure out a best placement so the nova scheduler would accept it IMHO.","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"30844e40150c33070d72e0dbca82b6621617b05d","unresolved":false,"context_lines":[{"line_number":105,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":106,"context_line":"resources the Nova Scheduler will quickly (depending of the numbers of"},{"line_number":107,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":108,"context_line":"the proposed action will be denied even if possible."},{"line_number":109,"context_line":""},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_6f8ac9c0","line":108,"in_reply_to":"3aaa91ec_2ef660f9","updated":"2016-06-30 01:11:04.000000000","message":"Can you please clarify what you mean by \u0027figure out best placement so the scheduler would accept it\u0027?\n\nThe problem is that there\u0027s no [good/maintainable] way that Watcher would know about all the filters (some of which could be out-of-tree) someone is using, so we\u0027d be left with a virtually non-maintainable solution.\n\nI understand Nova doesn\u0027t suggest using select_destinations() for a dry-run, but can I ask that we re-examine this?  We all realize it\u0027s inherently racey, but I think that\u0027s a concession we could make for the time being; the alternatives are much worse, IMO.\n\nThanks for your time.  :)\n\nThoughts?","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"f2e67e23764f23e884f6cd615920924784af647c","unresolved":false,"context_lines":[{"line_number":105,"context_line":"but never execute it and as the scheduler is virtually consuming the"},{"line_number":106,"context_line":"resources the Nova Scheduler will quickly (depending of the numbers of"},{"line_number":107,"context_line":"available hosts) raise the NoValidHost exception. The consequence is that"},{"line_number":108,"context_line":"the proposed action will be denied even if possible."},{"line_number":109,"context_line":""},{"line_number":110,"context_line":""},{"line_number":111,"context_line":"We perform some benchmark in order to evaluate if the nova-scheduler"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_661dc461","line":108,"in_reply_to":"3aaa91ec_6f8ac9c0","updated":"2016-06-30 14:57:14.000000000","message":"[Filters out of tree]\u003d\u003e I agree, this is a limitation of this approach. Do you think this is something that happen often ? \n\nWe could integrate watcher in an Openstack distribution with a list of known filters. We could also integrate the filters out-of-tree through plugins in the strategies. However, this solution is highly dependent on the techniques used by the strategies. \nI agree that dry-run can be a first approach but in a long term this is not sustainable, IMO.","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"dde9ae3d2e915dbee256612ab39c053f9b9dcbdc","unresolved":false,"context_lines":[{"line_number":169,"context_line":""},{"line_number":170,"context_line":"As shown in the diagram above, the `Watcher Decision Engine`_ service should"},{"line_number":171,"context_line":"fetch the Nova policies in order to be aware of the instance placement"},{"line_number":172,"context_line":"constraints."},{"line_number":173,"context_line":""},{"line_number":174,"context_line":"The selected strategy is able then to compute the needed actions for the"},{"line_number":175,"context_line":"solution in order to achieve its goals without violating the Nova constraints."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_8ece7481","line":172,"updated":"2016-06-28 13:55:57.000000000","message":"That\u0027s about maybe having a discovery API... Mmmm, probably, but we should be discussing about that later.","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":19214,"name":"Vincent Mahe","email":"vincent.mahe@b-com.com","username":"vmahe35"},"change_message_id":"b4a22e94e5e398f6151c9b8208bb3cab02707463","unresolved":false,"context_lines":[{"line_number":254,"context_line":""},{"line_number":255,"context_line":"It would be less intrusive if Nova could publish some events on the bus each"},{"line_number":256,"context_line":"time a new policy/constraint is added to Nova. This \"event sourcing\" approach"},{"line_number":257,"context_line":"would enable other components (such as Watcher or Congress) to subscribe to"},{"line_number":258,"context_line":"these events and use them to trigger other events or build their own structural"},{"line_number":259,"context_line":"model for their own needs."},{"line_number":260,"context_line":"It would also enable the developers to improve their optimization strategies"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_c3b151e3","line":257,"range":{"start_line":257,"start_character":19,"end_line":257,"end_character":29},"updated":"2016-06-23 16:03:21.000000000","message":"Openstack services","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"5545122c28ac0a6c1a45d4ea7b55fc2fed4a976b","unresolved":false,"context_lines":[{"line_number":254,"context_line":""},{"line_number":255,"context_line":"It would be less intrusive if Nova could publish some events on the bus each"},{"line_number":256,"context_line":"time a new policy/constraint is added to Nova. This \"event sourcing\" approach"},{"line_number":257,"context_line":"would enable other components (such as Watcher or Congress) to subscribe to"},{"line_number":258,"context_line":"these events and use them to trigger other events or build their own structural"},{"line_number":259,"context_line":"model for their own needs."},{"line_number":260,"context_line":"It would also enable the developers to improve their optimization strategies"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_81548554","line":257,"range":{"start_line":257,"start_character":19,"end_line":257,"end_character":29},"in_reply_to":"3aaa91ec_c3b151e3","updated":"2016-06-29 06:45:03.000000000","message":"Done","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":19214,"name":"Vincent Mahe","email":"vincent.mahe@b-com.com","username":"vmahe35"},"change_message_id":"b4a22e94e5e398f6151c9b8208bb3cab02707463","unresolved":false,"context_lines":[{"line_number":258,"context_line":"these events and use them to trigger other events or build their own structural"},{"line_number":259,"context_line":"model for their own needs."},{"line_number":260,"context_line":"It would also enable the developers to improve their optimization strategies"},{"line_number":261,"context_line":"by rebuilding Nova context from the event store."},{"line_number":262,"context_line":"We think scheduling needs to become more event and message driven but this will"},{"line_number":263,"context_line":"required to reconsider the OpenStack Architecture."},{"line_number":264,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_a302456d","line":261,"range":{"start_line":261,"start_character":47,"end_line":261,"end_character":48},"updated":"2016-06-23 16:03:21.000000000","message":"this would also make sense for other resource schedulers such as Neutron, Cinder, ...","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"5545122c28ac0a6c1a45d4ea7b55fc2fed4a976b","unresolved":false,"context_lines":[{"line_number":258,"context_line":"these events and use them to trigger other events or build their own structural"},{"line_number":259,"context_line":"model for their own needs."},{"line_number":260,"context_line":"It would also enable the developers to improve their optimization strategies"},{"line_number":261,"context_line":"by rebuilding Nova context from the event store."},{"line_number":262,"context_line":"We think scheduling needs to become more event and message driven but this will"},{"line_number":263,"context_line":"required to reconsider the OpenStack Architecture."},{"line_number":264,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_c14e8dc0","line":261,"range":{"start_line":261,"start_character":47,"end_line":261,"end_character":48},"in_reply_to":"3aaa91ec_a302456d","updated":"2016-06-29 06:45:03.000000000","message":"I agree","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":19214,"name":"Vincent Mahe","email":"vincent.mahe@b-com.com","username":"vmahe35"},"change_message_id":"b4a22e94e5e398f6151c9b8208bb3cab02707463","unresolved":false,"context_lines":[{"line_number":261,"context_line":"by rebuilding Nova context from the event store."},{"line_number":262,"context_line":"We think scheduling needs to become more event and message driven but this will"},{"line_number":263,"context_line":"required to reconsider the OpenStack Architecture."},{"line_number":264,"context_line":""},{"line_number":265,"context_line":"Data model impact"},{"line_number":266,"context_line":"-----------------"},{"line_number":267,"context_line":"None"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3aaa91ec_43ca816e","line":264,"updated":"2016-06-23 16:03:21.000000000","message":"This even more confirmed by the Use Cases listed regarding Nova events for SystemUsageData:\nhttps://wiki.openstack.org/wiki/SystemUsageData#User_stories\nwhere it says:\n\"As a systems integrator or enterprise cloud administrator, I want to monitor usage data over time so that I can optimize the utilization of my resources.\"","commit_id":"899ac1a40cfdaf692f4b57063382614fa6995cf5"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":27,"context_line":"    when there are many constraints ?"},{"line_number":28,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":29,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":30,"context_line":"    group that meets any constraints ?"},{"line_number":31,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":32,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":33,"context_line":"    used ?"}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_02a840da","line":30,"updated":"2016-07-07 07:15:47.000000000","message":"Shouldn\u0027t be \"all constraints\"?","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":27,"context_line":"    when there are many constraints ?"},{"line_number":28,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":29,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":30,"context_line":"    group that meets any constraints ?"},{"line_number":31,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":32,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":33,"context_line":"    used ?"}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_3c92ef4b","line":30,"in_reply_to":"1aa78d24_02a840da","updated":"2016-07-13 13:19:47.000000000","message":"yes","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":41,"context_line":"(holistic approach) which has a similar objective as `super-scheduler`_ in a"},{"line_number":42,"context_line":"long term."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Watcher doesn\u0027t want to check every potential placements"},{"line_number":45,"context_line":"with the Nova Scheduler before making a choice. This approach requires a huge"},{"line_number":46,"context_line":"number of queries and it may never be able to find a viable solution."},{"line_number":47,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_8131a15c","line":44,"range":{"start_line":44,"start_character":0,"end_line":44,"end_character":56},"updated":"2016-07-07 07:15:47.000000000","message":"There is one more problem as far as I know: some large VMs cannot be migrated because of their size (Nova can\u0027t migrate the largest instances). But Watcher might still want to move them. Is that covered by this blueprint or select-destinations-filter blueprint?","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":41,"context_line":"(holistic approach) which has a similar objective as `super-scheduler`_ in a"},{"line_number":42,"context_line":"long term."},{"line_number":43,"context_line":""},{"line_number":44,"context_line":"Watcher doesn\u0027t want to check every potential placements"},{"line_number":45,"context_line":"with the Nova Scheduler before making a choice. This approach requires a huge"},{"line_number":46,"context_line":"number of queries and it may never be able to find a viable solution."},{"line_number":47,"context_line":""}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_05d2bd5b","line":44,"range":{"start_line":44,"start_character":0,"end_line":44,"end_character":56},"in_reply_to":"1aa78d24_8131a15c","updated":"2016-07-13 13:19:47.000000000","message":"IMHO, this is a nova problem. However, we can maybe take into account this problem.","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":205,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":206,"context_line":"(Watcher will automatically apply the Audit)."},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"It will be also be usefull to have the `specify-multiple-live-migrate-hosts`_"},{"line_number":209,"context_line":"blueprint implemented to directly propose to nova-scheduler"},{"line_number":210,"context_line":"several destinations that meets the tenants constraints at the moment"},{"line_number":211,"context_line":"where they have been computed by the strategies."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_06d027f5","line":208,"range":{"start_line":208,"start_character":8,"end_line":208,"end_character":18},"updated":"2016-07-07 07:15:47.000000000","message":"Please remove one \"be\".","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":205,"context_line":"blueprint which aims to reduce the time between the audit and his application"},{"line_number":206,"context_line":"(Watcher will automatically apply the Audit)."},{"line_number":207,"context_line":""},{"line_number":208,"context_line":"It will be also be usefull to have the `specify-multiple-live-migrate-hosts`_"},{"line_number":209,"context_line":"blueprint implemented to directly propose to nova-scheduler"},{"line_number":210,"context_line":"several destinations that meets the tenants constraints at the moment"},{"line_number":211,"context_line":"where they have been computed by the strategies."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_85c6ad97","line":208,"range":{"start_line":208,"start_character":8,"end_line":208,"end_character":18},"in_reply_to":"1aa78d24_06d027f5","updated":"2016-07-13 13:19:47.000000000","message":"+1","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":215,"context_line":"in order to pro-actively check the validity of the whole Audit."},{"line_number":216,"context_line":"After this step if the Audit is not stale we can be confident for his"},{"line_number":217,"context_line":"application."},{"line_number":218,"context_line":"However, depending of the number of Actions that contains the"},{"line_number":219,"context_line":"Audit this application can take time."},{"line_number":220,"context_line":"So, we will verify again atomically the Action. In this way, the code of the"},{"line_number":221,"context_line":"Action `migration`_ will be amended to call the `select_destinations`_ method"},{"line_number":222,"context_line":"just prior the execution."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_01cdb184","line":219,"range":{"start_line":218,"start_character":0,"end_line":219,"end_character":37},"updated":"2016-07-07 07:15:47.000000000","message":"I think that the race condition is unavoidable. Of course, multiple checks could be done in every step (in strategy, applier, action), but in the end of the day Nova is the ultimate \"source of truth\" so maybe it\u0027s not worth doing it multiple times?\n\nI don\u0027t know how costly it is to get updates on Nova policies each time, but it will likely decrease the Watcher performance. Is it very likely that after the first check (in strategy), the policy will change so the initial plan cannot be executed? For that to happen an administrator would have to sit and change these policies all the time. Is this the typical scenario?\n\nI would rather optimize for typical scenario, do the policy check only once and simply accept the fail from Nova in these rare cases when the policy has changed.","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":215,"context_line":"in order to pro-actively check the validity of the whole Audit."},{"line_number":216,"context_line":"After this step if the Audit is not stale we can be confident for his"},{"line_number":217,"context_line":"application."},{"line_number":218,"context_line":"However, depending of the number of Actions that contains the"},{"line_number":219,"context_line":"Audit this application can take time."},{"line_number":220,"context_line":"So, we will verify again atomically the Action. In this way, the code of the"},{"line_number":221,"context_line":"Action `migration`_ will be amended to call the `select_destinations`_ method"},{"line_number":222,"context_line":"just prior the execution."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_65cbc172","line":219,"range":{"start_line":218,"start_character":0,"end_line":219,"end_character":37},"in_reply_to":"1aa78d24_01cdb184","updated":"2016-07-13 13:19:47.000000000","message":"I don\u0027t want to address here the problem of the performances. \n\nI want to focus on \u003d\u003e How watcher strategies should\ntake into account the nova policies \n\nThe performance can be addressed later.","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":223,"context_line":""},{"line_number":224,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":225,"context_line":"will be amended to add the get_policies method. We should adapt all existing"},{"line_number":226,"context_line":"strategies so that they take into account the Nova constraints."},{"line_number":227,"context_line":""},{"line_number":228,"context_line":"We shoud provide a helper for the strategies developer to query the nova"},{"line_number":229,"context_line":"policies."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_a698331d","line":226,"updated":"2016-07-07 07:15:47.000000000","message":"Just an idea: this change will ENABLE strategies to adopt to Nova policies. What about FORCING it? For example there could be one more step in base strategy execute method like \"verify_solution\", which could throw exception or something if a strategy produced actions which cannot be applied.","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":223,"context_line":""},{"line_number":224,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":225,"context_line":"will be amended to add the get_policies method. We should adapt all existing"},{"line_number":226,"context_line":"strategies so that they take into account the Nova constraints."},{"line_number":227,"context_line":""},{"line_number":228,"context_line":"We shoud provide a helper for the strategies developer to query the nova"},{"line_number":229,"context_line":"policies."}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_e54e9194","line":226,"in_reply_to":"1aa78d24_a698331d","updated":"2016-07-13 13:19:47.000000000","message":"If I understand well your assumption. This is the same approach as \"select_destinations\u0027 and I don\u0027t think this is a good idea.\n\u003d\u003e https://blueprints.launchpad.net/nova/+spec/specify-multiple-live-migrate-hosts","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":20515,"name":"Tomasz Kaczynski","email":"tomasz.kaczynski@intel.com","username":"tkaczynski"},"change_message_id":"4a43fbe5d2d7466a2ff5846e41b3ee6b740c0e6d","unresolved":false,"context_lines":[{"line_number":230,"context_line":""},{"line_number":231,"context_line":"Elements that should be discussed with the Nova team:"},{"line_number":232,"context_line":""},{"line_number":233,"context_line":"We need to relies on a API that won\u0027t exist for some time."},{"line_number":234,"context_line":"This spec will need that the `select_destinations`_ method to be exposed"},{"line_number":235,"context_line":"as a RESTFul API."},{"line_number":236,"context_line":"This spec will also need  that the `RequestSpec`_ object to be exposed as"}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_05e04af9","line":233,"range":{"start_line":233,"start_character":11,"end_line":233,"end_character":17},"updated":"2016-07-07 07:15:47.000000000","message":"rely","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"b32f0123d0c09d428be1976bb19d024e5f5a203a","unresolved":false,"context_lines":[{"line_number":230,"context_line":""},{"line_number":231,"context_line":"Elements that should be discussed with the Nova team:"},{"line_number":232,"context_line":""},{"line_number":233,"context_line":"We need to relies on a API that won\u0027t exist for some time."},{"line_number":234,"context_line":"This spec will need that the `select_destinations`_ method to be exposed"},{"line_number":235,"context_line":"as a RESTFul API."},{"line_number":236,"context_line":"This spec will also need  that the `RequestSpec`_ object to be exposed as"}],"source_content_type":"text/x-rst","patch_set":8,"id":"1aa78d24_a5a649a8","line":233,"range":{"start_line":233,"start_character":11,"end_line":233,"end_character":17},"in_reply_to":"1aa78d24_05e04af9","updated":"2016-07-13 13:19:47.000000000","message":"+1","commit_id":"c0f8f3a474582ffd172f2c5e6dae952d01a9372f"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"df99be612310985e62c07a34d55a5fbb1dce1636","unresolved":false,"context_lines":[{"line_number":21,"context_line":"strategies take into account the nova policies ?"},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Open questions :"},{"line_number":24,"context_line":"  - Do we want Watcher to be able to optimize an infrastructure"},{"line_number":25,"context_line":"    when there are many constraints ?"},{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_d24f8268","line":24,"updated":"2016-07-09 02:57:34.000000000","message":"Yes","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":6437,"name":"Susanne Balle","email":"sleipnir012@gmail.com","username":"susanne-balle"},"change_message_id":"af8280def1ac0c19aa2bb1d83fa16db93a19c047","unresolved":false,"context_lines":[{"line_number":21,"context_line":"strategies take into account the nova policies ?"},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Open questions :"},{"line_number":24,"context_line":"  - Do we want Watcher to be able to optimize an infrastructure"},{"line_number":25,"context_line":"    when there are many constraints ?"},{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_b0b9207c","line":24,"in_reply_to":"1aa78d24_c53275f1","updated":"2016-07-14 15:24:18.000000000","message":"So I agree and will move these questions down to the bottom given that they are out of place.","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"e750e2f40a4cf0e2de700b1e292a454715cb97b0","unresolved":false,"context_lines":[{"line_number":21,"context_line":"strategies take into account the nova policies ?"},{"line_number":22,"context_line":""},{"line_number":23,"context_line":"Open questions :"},{"line_number":24,"context_line":"  - Do we want Watcher to be able to optimize an infrastructure"},{"line_number":25,"context_line":"    when there are many constraints ?"},{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_c53275f1","line":24,"in_reply_to":"1aa78d24_d24f8268","updated":"2016-07-13 13:10:49.000000000","message":"+1","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"df99be612310985e62c07a34d55a5fbb1dce1636","unresolved":false,"context_lines":[{"line_number":23,"context_line":"Open questions :"},{"line_number":24,"context_line":"  - Do we want Watcher to be able to optimize an infrastructure"},{"line_number":25,"context_line":"    when there are many constraints ?"},{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_3257e603","line":26,"updated":"2016-07-09 02:57:34.000000000","message":"Yes :-)","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":6437,"name":"Susanne Balle","email":"sleipnir012@gmail.com","username":"susanne-balle"},"change_message_id":"af8280def1ac0c19aa2bb1d83fa16db93a19c047","unresolved":false,"context_lines":[{"line_number":23,"context_line":"Open questions :"},{"line_number":24,"context_line":"  - Do we want Watcher to be able to optimize an infrastructure"},{"line_number":25,"context_line":"    when there are many constraints ?"},{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_f0aba81c","line":26,"in_reply_to":"1aa78d24_3257e603","updated":"2016-07-14 15:24:18.000000000","message":"So I agree and will move these questions down to the bottom given that they are out of place.","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"df99be612310985e62c07a34d55a5fbb1dce1636","unresolved":false,"context_lines":[{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_72416e39","line":29,"updated":"2016-07-09 02:57:34.000000000","message":"Ideally yes, but not initially required - also complex to do this since migrations aren\u0027t transactional","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"e750e2f40a4cf0e2de700b1e292a454715cb97b0","unresolved":false,"context_lines":[{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_a4236fa6","line":29,"in_reply_to":"1aa78d24_72416e39","updated":"2016-07-13 13:10:49.000000000","message":"IMHO, we can have two type of affinity constraint. \n\nFirst, we only consider the model obtained as the end of the reconfiguration process. ( we can temporaly accept a violation)\n\nSecond, the constraint cannot be violated during the reconfiguration process and at the end.\n\nWhat do you think ?","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":19273,"name":"Fabien Hermenier","email":"fabien.hermenier@unice.fr","username":"fhermeni"},"change_message_id":"8a08f83c50d54cf98d1103bb15f14f484a1c3a9d","unresolved":false,"context_lines":[{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"fad55976_af371de2","line":29,"in_reply_to":"1aa78d24_9bfd250b","updated":"2016-07-15 23:09:34.000000000","message":"Jean-Emile might talk about discrete vs. continuous constraints ?\n\nWith a \"discrete\" anti-affinity, the constraint must only be satisfied by the end of the reconfiguration. That means it is ok to have both VMs co-located temporarily during the reconfiguration.\n\nUsing a \"continuous\" anti-affinity, this temporary violation is denied. \n\nSee \"Higher SLA Satisfaction in Datacenters\nwith Continuous VM Placement Constraints\" at HotDep 2013 (https://goo.gl/gkv6Nz). The slides: https://goo.gl/3OogeD","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":6437,"name":"Susanne Balle","email":"sleipnir012@gmail.com","username":"susanne-balle"},"change_message_id":"af8280def1ac0c19aa2bb1d83fa16db93a19c047","unresolved":false,"context_lines":[{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_9bfd250b","line":29,"in_reply_to":"1aa78d24_a4236fa6","updated":"2016-07-14 15:24:18.000000000","message":"To me a VM with an affinity rule can be migrated but still has to obey by the constraints. I am not sure what temporally accept a violation mean?","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"aff9f734bf8cf65e3e892ade1066b9da78e81ef1","unresolved":false,"context_lines":[{"line_number":26,"context_line":"  - Do we want Watcher to select instances to be moved, and also a preferred"},{"line_number":27,"context_line":"    set of hosts to move it to, and let Nova Scheduler pick a host from that"},{"line_number":28,"context_line":"    group that meets all constraints ?"},{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"}],"source_content_type":"text/x-rst","patch_set":14,"id":"dada55a8_217cf971","line":29,"in_reply_to":"fad55976_af371de2","updated":"2016-07-18 06:58:09.000000000","message":"yes ! thanks Fabien.","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"df99be612310985e62c07a34d55a5fbb1dce1636","unresolved":false,"context_lines":[{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"},{"line_number":33,"context_line":"  - Do we have to organize a cross project meeting ?"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_92469a4d","line":32,"updated":"2016-07-09 02:57:34.000000000","message":"I hope not","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"e750e2f40a4cf0e2de700b1e292a454715cb97b0","unresolved":false,"context_lines":[{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"},{"line_number":33,"context_line":"  - Do we have to organize a cross project meeting ?"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_a5164946","line":32,"in_reply_to":"1aa78d24_92469a4d","updated":"2016-07-13 13:10:49.000000000","message":"+1","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":6437,"name":"Susanne Balle","email":"sleipnir012@gmail.com","username":"susanne-balle"},"change_message_id":"af8280def1ac0c19aa2bb1d83fa16db93a19c047","unresolved":false,"context_lines":[{"line_number":29,"context_line":"  - Do we think that an instance with an affinity rule can be migrated ?"},{"line_number":30,"context_line":"    How often in an typical infrastructure deployment the affinity rule is"},{"line_number":31,"context_line":"    used ?"},{"line_number":32,"context_line":"  - Do we need to provide our own driver of the nova-scheduler ?"},{"line_number":33,"context_line":"  - Do we have to organize a cross project meeting ?"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_db55edfa","line":32,"in_reply_to":"1aa78d24_a5164946","updated":"2016-07-14 15:24:18.000000000","message":"I am assuming this will be clarified once we talk to he nova team.","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":7664,"name":"Joe Cropper","email":"jwcroppe@us.ibm.com","username":"jwcroppe"},"change_message_id":"df99be612310985e62c07a34d55a5fbb1dce1636","unresolved":false,"context_lines":[{"line_number":127,"context_line":"that it cannot integrate these requirements into the strategies themselves."},{"line_number":128,"context_line":"Watcher can only try a new placement for an instance and get it validated"},{"line_number":129,"context_line":"by the Nova Scheduler. If an instance has an affinity constraint, Watcher"},{"line_number":130,"context_line":"can try many combinations and never find a valid host. The right approach"},{"line_number":131,"context_line":"for this problem is to let Watcher fetch a snapshot of the Nova placement"},{"line_number":132,"context_line":"policies and try the host validation combinations internally."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"The fourth and last issue is that the `select_destinations`_  method cannot"},{"line_number":135,"context_line":"be used to validate instances with an affinity group. The"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_3241a60c","line":132,"range":{"start_line":130,"start_character":55,"end_line":132,"end_character":61},"updated":"2016-07-09 02:57:34.000000000","message":"How would this actually work?","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"c1314411510d9d24b12bc0702ab292ee4f38d0fe","unresolved":false,"context_lines":[{"line_number":127,"context_line":"that it cannot integrate these requirements into the strategies themselves."},{"line_number":128,"context_line":"Watcher can only try a new placement for an instance and get it validated"},{"line_number":129,"context_line":"by the Nova Scheduler. If an instance has an affinity constraint, Watcher"},{"line_number":130,"context_line":"can try many combinations and never find a valid host. The right approach"},{"line_number":131,"context_line":"for this problem is to let Watcher fetch a snapshot of the Nova placement"},{"line_number":132,"context_line":"policies and try the host validation combinations internally."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"The fourth and last issue is that the `select_destinations`_  method cannot"},{"line_number":135,"context_line":"be used to validate instances with an affinity group. The"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_a97583fa","line":132,"range":{"start_line":130,"start_character":55,"end_line":132,"end_character":61},"in_reply_to":"1aa78d24_3241a60c","updated":"2016-07-13 13:30:22.000000000","message":"fetch a snapshot or validate the combinations ?","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":6437,"name":"Susanne Balle","email":"sleipnir012@gmail.com","username":"susanne-balle"},"change_message_id":"af8280def1ac0c19aa2bb1d83fa16db93a19c047","unresolved":false,"context_lines":[{"line_number":127,"context_line":"that it cannot integrate these requirements into the strategies themselves."},{"line_number":128,"context_line":"Watcher can only try a new placement for an instance and get it validated"},{"line_number":129,"context_line":"by the Nova Scheduler. If an instance has an affinity constraint, Watcher"},{"line_number":130,"context_line":"can try many combinations and never find a valid host. The right approach"},{"line_number":131,"context_line":"for this problem is to let Watcher fetch a snapshot of the Nova placement"},{"line_number":132,"context_line":"policies and try the host validation combinations internally."},{"line_number":133,"context_line":""},{"line_number":134,"context_line":"The fourth and last issue is that the `select_destinations`_  method cannot"},{"line_number":135,"context_line":"be used to validate instances with an affinity group. The"}],"source_content_type":"text/x-rst","patch_set":14,"id":"1aa78d24_9b80854b","line":132,"range":{"start_line":130,"start_character":55,"end_line":132,"end_character":61},"in_reply_to":"1aa78d24_a97583fa","updated":"2016-07-14 15:24:18.000000000","message":"From the DB? I like the snapshot idea and I do not believe that existing policies/constraints associated with running VMs will change too much over time.","commit_id":"9c6d1575e156e84b6eddf4a85daf40d14f7258eb"},{"author":{"_account_id":11750,"name":"Antoine Cabot","email":"antoine.cabot@b-com.com","username":"antoinecabot"},"change_message_id":"5189e2551d41961ac446c7f2e21bb66b166b567a","unresolved":false,"context_lines":[{"line_number":73,"context_line":"In Watcher the two steps will be carried out independently. The strategy"},{"line_number":74,"context_line":"developer can create an audit encompassing a large number of VM placement"},{"line_number":75,"context_line":"algorithms including a combination of exact approaches and Heuristics"},{"line_number":76,"context_line":"or Meta-heuristics. As part of the decision phase (Wather strategy) Watcher"},{"line_number":77,"context_line":"needs to check the validity of configurations and/or test an optimization"},{"line_number":78,"context_line":"algorithm before applying it to the system. In some cases, a strategy can"},{"line_number":79,"context_line":"generate several possible solutions and uses an objective function to assess the"}],"source_content_type":"text/x-rst","patch_set":16,"id":"dada55a8_2b5666d2","line":76,"range":{"start_line":76,"start_character":51,"end_line":76,"end_character":57},"updated":"2016-07-19 00:26:55.000000000","message":"s/Wather/Watcher","commit_id":"e16883a491a4030e304319b7b2fe79226d3376d3"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"64aa7124715b62c2f9590c48fd376a59935780a5","unresolved":false,"context_lines":[{"line_number":73,"context_line":"In Watcher the two steps will be carried out independently. The strategy"},{"line_number":74,"context_line":"developer can create an audit encompassing a large number of VM placement"},{"line_number":75,"context_line":"algorithms including a combination of exact approaches and Heuristics"},{"line_number":76,"context_line":"or Meta-heuristics. As part of the decision phase (Wather strategy) Watcher"},{"line_number":77,"context_line":"needs to check the validity of configurations and/or test an optimization"},{"line_number":78,"context_line":"algorithm before applying it to the system. In some cases, a strategy can"},{"line_number":79,"context_line":"generate several possible solutions and uses an objective function to assess the"}],"source_content_type":"text/x-rst","patch_set":16,"id":"dada55a8_3d01156b","line":76,"range":{"start_line":76,"start_character":51,"end_line":76,"end_character":57},"in_reply_to":"dada55a8_2b5666d2","updated":"2016-07-20 23:04:46.000000000","message":"Done","commit_id":"e16883a491a4030e304319b7b2fe79226d3376d3"},{"author":{"_account_id":11750,"name":"Antoine Cabot","email":"antoine.cabot@b-com.com","username":"antoinecabot"},"change_message_id":"5189e2551d41961ac446c7f2e21bb66b166b567a","unresolved":false,"context_lines":[{"line_number":225,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":226,"context_line":"will be amended to add the get_policies method. All existing strategies should"},{"line_number":227,"context_line":"be modified to take the Nova constraints into account. This change will enable"},{"line_number":228,"context_line":"strategies to adher to the Nova policies. An additional step is to enfore the"},{"line_number":229,"context_line":"compliance by having the BaseStrategy class execute a \"verify_solution\" method"},{"line_number":230,"context_line":"and throw exception if a strategy produces actions which cannot be applied."},{"line_number":231,"context_line":""}],"source_content_type":"text/x-rst","patch_set":16,"id":"dada55a8_2bae6685","line":228,"range":{"start_line":228,"start_character":14,"end_line":228,"end_character":19},"updated":"2016-07-19 00:26:55.000000000","message":"s/adher/adhere","commit_id":"e16883a491a4030e304319b7b2fe79226d3376d3"},{"author":{"_account_id":16495,"name":"Jean-Emile DARTOIS","email":"jean-emile.dartois@b-com.com","username":"JeD"},"change_message_id":"64aa7124715b62c2f9590c48fd376a59935780a5","unresolved":false,"context_lines":[{"line_number":225,"context_line":"The BaseStrategy class watcher/decision_engine/strategy/strategies/base.py"},{"line_number":226,"context_line":"will be amended to add the get_policies method. All existing strategies should"},{"line_number":227,"context_line":"be modified to take the Nova constraints into account. This change will enable"},{"line_number":228,"context_line":"strategies to adher to the Nova policies. An additional step is to enfore the"},{"line_number":229,"context_line":"compliance by having the BaseStrategy class execute a \"verify_solution\" method"},{"line_number":230,"context_line":"and throw exception if a strategy produces actions which cannot be applied."},{"line_number":231,"context_line":""}],"source_content_type":"text/x-rst","patch_set":16,"id":"dada55a8_7df70d8b","line":228,"range":{"start_line":228,"start_character":14,"end_line":228,"end_character":19},"in_reply_to":"dada55a8_2bae6685","updated":"2016-07-20 23:04:46.000000000","message":"Done","commit_id":"e16883a491a4030e304319b7b2fe79226d3376d3"}]}
