)]}'
{"controller.rst":[{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":28,"context_line":"* Processing Amphora configuration updates and making them available to the"},{"line_number":29,"context_line":"  Amphora driver"},{"line_number":30,"context_line":"* Providing certificate information to the Amphora driver"},{"line_number":31,"context_line":"* Building Amphora instances"},{"line_number":32,"context_line":"* Managing the Amphora spares pool"},{"line_number":33,"context_line":"* Cleaning up Amphora instances that are no longer needed"},{"line_number":34,"context_line":"* Monitoring the health of Amphora instances"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_b3d62052","line":31,"updated":"2014-12-05 04:30:08.000000000","message":"I would say \"Deploy Amphora Instances\" since \"Building\" is a word that has connotations around compiling and/or creating a master image.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":28,"context_line":"* Processing Amphora configuration updates and making them available to the"},{"line_number":29,"context_line":"  Amphora driver"},{"line_number":30,"context_line":"* Providing certificate information to the Amphora driver"},{"line_number":31,"context_line":"* Building Amphora instances"},{"line_number":32,"context_line":"* Managing the Amphora spares pool"},{"line_number":33,"context_line":"* Cleaning up Amphora instances that are no longer needed"},{"line_number":34,"context_line":"* Monitoring the health of Amphora instances"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_e768de8a","line":31,"in_reply_to":"5a890539_b3d62052","updated":"2014-12-05 08:00:24.000000000","message":"Booting? That\u0027s the verb Nova uses. Or, \"Managing Amphorae Lifecycle\" would wrap this and the next two below into a single point, though it\u0027s more explicit with them separate.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":28,"context_line":"* Processing Amphora configuration updates and making them available to the"},{"line_number":29,"context_line":"  Amphora driver"},{"line_number":30,"context_line":"* Providing certificate information to the Amphora driver"},{"line_number":31,"context_line":"* Building Amphora instances"},{"line_number":32,"context_line":"* Managing the Amphora spares pool"},{"line_number":33,"context_line":"* Cleaning up Amphora instances that are no longer needed"},{"line_number":34,"context_line":"* Monitoring the health of Amphora instances"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_0717aa46","line":31,"in_reply_to":"5a890539_b3d62052","updated":"2014-12-05 07:33:46.000000000","message":"we should probably change the compute interface\u0027s method of build to deploy.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":35,"context_line":"* Processing alerts and messages from the Amphora (example \"member down\")"},{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_8759da37","line":38,"updated":"2014-12-05 07:33:46.000000000","message":"I don\u0027t agree with stating Ceilometer specifically.  I think calling it a \"metering service\" would be fine as different operators will use different metering services.  I know we are going to use Stacktach, and I\u0027ve been told that HP might as well, or at least yall should look into it because there are HP people working on it.  Ceilometer has scaling issues.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":35,"context_line":"* Processing alerts and messages from the Amphora (example \"member down\")"},{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_47d5321a","line":38,"updated":"2014-12-05 08:00:24.000000000","message":"I wonder if it is possible to have Amphorae send their data directly to Ceilometer/metering, reducing the Controller workload to make scaling up the number of Amphorae::Controller easier... If people think that\u0027s crazy, at least. :)","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"8e21ea28f87cbc34c434b46735cea9fb7f987620","unresolved":false,"context_lines":[{"line_number":35,"context_line":"* Processing alerts and messages from the Amphora (example \"member down\")"},{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_e7460b7b","line":38,"in_reply_to":"5a890539_47d5321a","updated":"2014-12-05 17:15:20.000000000","message":"we need to do some massaging of the data, and I\u0027m not sure if any of the metering sevices can do custom logic.  plus, I\u0027m not sure if the metering services will be able scale to large amounts of amphorae connecting to its service every poll interval.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":35,"context_line":"* Processing alerts and messages from the Amphora (example \"member down\")"},{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_ebf10d02","line":38,"in_reply_to":"5a890539_e7460b7b","updated":"2014-12-11 00:12:42.000000000","message":"+1 and we have a concept that our management/metering traffic flows over the \"load balancer\" network which is not intended to have routers.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"9974ccd3f570ce975faa64f0b0c117b195f8c84a","unresolved":false,"context_lines":[{"line_number":35,"context_line":"* Processing alerts and messages from the Amphora (example \"member down\")"},{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_da32b85d","line":38,"in_reply_to":"5a890539_e7460b7b","updated":"2014-12-06 00:37:48.000000000","message":"https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-meter-lbaas is that blueprint we have in mind","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"},{"line_number":42,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_c78e8232","line":39,"updated":"2014-12-05 08:00:24.000000000","message":"I think the actual API is going to be running as a separate instance, and possibly even on separate nodes? API -\u003e DB / Queue -\u003e Controller\n\nUnless what you mean is that it\u0027ll have to \"do work in response to\" API requests (note the passive voice / disconnected implication)? I think that\u0027s possible from what I read in the API section below, but would like explicit confirmation.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"},{"line_number":42,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_6b535de2","line":39,"in_reply_to":"5a890539_1a0d2e36","updated":"2014-12-11 00:12:42.000000000","message":"Yes, my intention here is that the API is running as it\u0027s own process outside of the scope of the controller.  This is the \"controller\" part that pulls requests from the queue/messaging bus.\nWe have also talked about not allowing the API server direct Octavia database access, so this component would fulfill query requests.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"8e21ea28f87cbc34c434b46735cea9fb7f987620","unresolved":false,"context_lines":[{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"},{"line_number":42,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_1a0d2e36","line":39,"in_reply_to":"5a890539_c78e8232","updated":"2014-12-05 17:15:20.000000000","message":"good catch.  The octavia-api will be a separate process and it will respond to any API request and forward any CUD operations to the controller via Queue","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Proposed change"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_b3bd4006","line":40,"updated":"2014-12-05 04:30:08.000000000","message":"I really hope we can send these outside the purview of the controller. But I realize we still need to figure that out. And there are probably other amphora data that will need to be proxied anyway.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  Ceilometer"},{"line_number":39,"context_line":"* Responding to API requests"},{"line_number":40,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":41,"context_line":"  archival"},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Proposed change"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_4bcdc184","line":40,"in_reply_to":"5a890539_b3bd4006","updated":"2014-12-11 00:12:42.000000000","message":"+1 - This is a placeholder.  We will likely need to proxy access to Swift to be able to store the logs in the customer\u0027s Swift container using the request\u0027s auth token.  Not sure we want to send that down to the Amp.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":57,"context_line":" .. image:: OctaviaController.png"},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"These components should be implemented as independent processes that"},{"line_number":60,"context_line":"communicate with each other via zeromq unless otherwise noted.  This provides"},{"line_number":61,"context_line":"a level of autonomy to these functions."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The highly available database will provide the persistent \"brain\" for the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_a79fb650","line":60,"updated":"2014-12-05 08:00:24.000000000","message":"Yeah, we\u0027ve been referring to these processes as \"agents\" in our discussions, but essentially that\u0027s what we\u0027re thinking too. A collection of agents/processes.\n\nIs there a more generic way to express this than with a specific software solution? I am reading about ZeroMQ now, but is it possible to do this with a similar solution, or without it entirely? It\u0027s just a specific implementation of inter-process communication, right?","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":57,"context_line":" .. image:: OctaviaController.png"},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"These components should be implemented as independent processes that"},{"line_number":60,"context_line":"communicate with each other via zeromq unless otherwise noted.  This provides"},{"line_number":61,"context_line":"a level of autonomy to these functions."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The highly available database will provide the persistent \"brain\" for the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_8b9a496c","line":60,"in_reply_to":"5a890539_7a1aa4c7","updated":"2014-12-11 00:12:42.000000000","message":"+1 Plus I wanted to be specific here to have the conversation.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"9974ccd3f570ce975faa64f0b0c117b195f8c84a","unresolved":false,"context_lines":[{"line_number":57,"context_line":" .. image:: OctaviaController.png"},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"These components should be implemented as independent processes that"},{"line_number":60,"context_line":"communicate with each other via zeromq unless otherwise noted.  This provides"},{"line_number":61,"context_line":"a level of autonomy to these functions."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The highly available database will provide the persistent \"brain\" for the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_7a1aa4c7","line":60,"in_reply_to":"5a890539_a79fb650","updated":"2014-12-06 00:37:48.000000000","message":"We said zeroMQ because we weren\u0027t prepared to run oslo messaging; we might laso be able to use taskflow for some stuff...","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":72,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":73,"context_line":"a given controller is primarily for distributing monitoring and network load."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"**Amphora Driver**"},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_f31348ff","line":73,"updated":"2014-12-05 04:30:08.000000000","message":"Ok, so we need to have a discussion about this, as I\u0027m not sure I agree here-- I suspect an \"anyone can talk to any amphora\" design is going to lead to some scaling issued one way or another. I agree that trying to minimize state to whatever is in the database, as well as making sure controllers are disposable is a really good idea, but there are other ways of going about it than what you\u0027ve suggested in this paragraph.\n\nIn any case, I\u0027m pretty sure we\u0027re not going to be able to agree on this one until we\u0027ve white-boarded everything out at the hack-a-thon in a little over a week.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"9974ccd3f570ce975faa64f0b0c117b195f8c84a","unresolved":false,"context_lines":[{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":72,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":73,"context_line":"a given controller is primarily for distributing monitoring and network load."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"**Amphora Driver**"},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_9ac4b00f","line":73,"in_reply_to":"5a890539_67578e5b","updated":"2014-12-06 00:37:48.000000000","message":"+1","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":72,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":73,"context_line":"a given controller is primarily for distributing monitoring and network load."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"**Amphora Driver**"},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_67578e5b","line":73,"in_reply_to":"5a890539_f31348ff","updated":"2014-12-05 08:00:24.000000000","message":"I am +1 to the proposal as is, sorry Stephen -- this is essentially what we\u0027ve been tossing around internally. Amphorae should not be tied to a specific controller, other than maybe for monitoring purposes. Any controller should be able to update configuration on any Amphorae at any time. I think that approach should actually help us AVOID scaling issues.\n\nHackathon ahoy!","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":72,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":73,"context_line":"a given controller is primarily for distributing monitoring and network load."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"**Amphora Driver**"},{"line_number":76,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_4784525d","line":73,"in_reply_to":"5a890539_f31348ff","updated":"2014-12-05 07:33:46.000000000","message":"Yeah this paragraph was something similar to what whiteboarded here.  I know you want more details on this and we\u0027re going to get them, hopefully before the hackathon.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":86,"context_line":"requests can be serviced by querying the database for the current state of"},{"line_number":87,"context_line":"load balancers and listeners."},{"line_number":88,"context_line":""},{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_9378442e","line":89,"updated":"2014-12-05 04:30:08.000000000","message":"\"may cause\" After all if the spares pool has enough amphorae in it, no need to build another.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":86,"context_line":"requests can be serviced by querying the database for the current state of"},{"line_number":87,"context_line":"load balancers and listeners."},{"line_number":88,"context_line":""},{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_4bd061f6","line":89,"in_reply_to":"5a890539_679cee0f","updated":"2014-12-11 00:12:42.000000000","message":"This has nothing to do with the spares pool.  The component formerly known as \"Build Manager\" is only responsible for wiring up AMP network ports/addresses, etc.  It does not build spares.\nThis process can be lengthy in the near future when we are wiring up multiple AMPs for HA and scalability.\nThe only time, that I can think of, it *may* need to trigger an AMP \"boot\" is if there is location dependency that the current spares pool/AMPs cannot meet or the spares pool is set to zero.\nI hope the name change helps clarify my intent for this component.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":86,"context_line":"requests can be serviced by querying the database for the current state of"},{"line_number":87,"context_line":"load balancers and listeners."},{"line_number":88,"context_line":""},{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_679cee0f","line":89,"in_reply_to":"5a890539_9378442e","updated":"2014-12-05 08:00:24.000000000","message":"Yes -- it should query the \"spares pool\" for an Amphora, and if there isn\u0027t one (ran out, or spare pool size is 0) THEN it could request one be built specifically. Otherwise, the spare pool manager agent/process should be monitoring the pool and building Amphorae.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"API requests may also cause the API Manager to push updated configurations to"},{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_07e16a2a","line":92,"updated":"2014-12-05 07:33:46.000000000","message":"If this isn\u0027t an agent running, I\u0027d suggest spawning a thread since the main thread and this spawned thread would be bottlenecked by network I/O, so python\u0027s GIL wouldn\u0027t get in the way.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"API requests may also cause the API Manager to push updated configurations to"},{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_42d2805c","line":92,"in_reply_to":"5a890539_07e16a2a","updated":"2014-12-05 08:00:24.000000000","message":"I think this would be part of the build agent and not spawned from the API Manager -- see my first comment below under Build Manager.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"API requests may also cause the API Manager to push updated configurations to"},{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_0b02d96e","line":92,"in_reply_to":"5a890539_3ad57cb7","updated":"2014-12-11 00:12:42.000000000","message":"My intent here is the API manager would spawn the component formerly known as \"Build Manager\".  There is no need for them to stick around and they will each potentially be focused on lengthy deployment flows.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"9974ccd3f570ce975faa64f0b0c117b195f8c84a","unresolved":false,"context_lines":[{"line_number":89,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"},{"line_number":90,"context_line":"spawn a Build Manager process.  Spawning a seperate process makes sure that"},{"line_number":91,"context_line":"the API Manager can continue to service API requests while the longer running"},{"line_number":92,"context_line":"build process is progressing."},{"line_number":93,"context_line":""},{"line_number":94,"context_line":"API requests may also cause the API Manager to push updated configurations to"},{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_3ad57cb7","line":92,"in_reply_to":"5a890539_42d2805c","updated":"2014-12-06 00:37:48.000000000","message":"+! That should be thrown on the queue and done in the \"build agent\"","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"},{"line_number":96,"context_line":"to the listener."},{"line_number":97,"context_line":""},{"line_number":98,"context_line":"Delete requests for load balancers or listeners will cause an delete call"},{"line_number":99,"context_line":"to the Amphora driver and a pending delete status set in the database for"},{"line_number":100,"context_line":"the Housekeeping Manager to cleanup the Amphora should it no longer be needed."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_73e918b7","line":98,"updated":"2014-12-05 04:30:08.000000000","message":"\"may cause\" Again, if the listener isn\u0027t the last one on a given loadbalancer or the loadbalancer instance isn\u0027t the last one on the amphora, we don\u0027t kill the amphora yet.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":95,"context_line":"the Octavia database and notify the driver to push an updated configuration"},{"line_number":96,"context_line":"to the listener."},{"line_number":97,"context_line":""},{"line_number":98,"context_line":"Delete requests for load balancers or listeners will cause an delete call"},{"line_number":99,"context_line":"to the Amphora driver and a pending delete status set in the database for"},{"line_number":100,"context_line":"the Housekeeping Manager to cleanup the Amphora should it no longer be needed."},{"line_number":101,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_8bede9b9","line":98,"in_reply_to":"5a890539_73e918b7","updated":"2014-12-11 00:12:42.000000000","message":"Yes, as stated at the end of the sentence.  I have clarified the wording in the next patch set.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":104,"context_line":"The Build Manager is a spawned process from the API Manager or the Health"},{"line_number":105,"context_line":"Manager.  It interfaces with the compute driver (in some deployment scenarios),"},{"line_number":106,"context_line":"network driver, and Amphora driver to activate Amphora instances,"},{"line_number":107,"context_line":"load balancers, and listeners."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_87b8fa76","line":107,"updated":"2014-12-05 08:00:24.000000000","message":"I think Build Manager and Housekeeping manager are a little bit mixed/intertwined -- it should be one process managing the lifecycle of the Amphorae.\n\n* Making sure the spares pool is filled\n* Generating new Amphorae on request if the spares pool is empty somehow or size zero\n* Deleting failed or otherwise marked for deletion Amphorae\n\nThus, this should be part of a full-time running agent.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":104,"context_line":"The Build Manager is a spawned process from the API Manager or the Health"},{"line_number":105,"context_line":"Manager.  It interfaces with the compute driver (in some deployment scenarios),"},{"line_number":106,"context_line":"network driver, and Amphora driver to activate Amphora instances,"},{"line_number":107,"context_line":"load balancers, and listeners."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_8b4409b8","line":107,"in_reply_to":"5a890539_87b8fa76","updated":"2014-12-11 00:12:42.000000000","message":"These are all activities the Housekeeping Manager handles and not the component formerly known as \"Build Manager\"","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":107,"context_line":"load balancers, and listeners."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"},{"line_number":111,"context_line":"Amphora via the network driver and triggering a configuration push via the"},{"line_number":112,"context_line":"Amphora driver."},{"line_number":113,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_07eb2add","line":110,"updated":"2014-12-05 07:33:46.000000000","message":"can\u0027t assume Neutron here, should be generic.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":107,"context_line":"load balancers, and listeners."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"},{"line_number":111,"context_line":"Amphora via the network driver and triggering a configuration push via the"},{"line_number":112,"context_line":"Amphora driver."},{"line_number":113,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_6b1efdbc","line":110,"in_reply_to":"5a890539_07eb2add","updated":"2014-12-11 00:12:42.000000000","message":"Yes, I have updated this to reflect the network driver in the next patch set.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"},{"line_number":111,"context_line":"Amphora via the network driver and triggering a configuration push via the"},{"line_number":112,"context_line":"Amphora driver."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"The Amphora configured by the Build Manager may be an existing Amphora"},{"line_number":115,"context_line":"instance or a new Amphora from the spares pool.  This determination will"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_07008a01","line":112,"updated":"2014-12-05 08:00:24.000000000","message":"This part of the build manager should just be part of the API Manager... As this would be entirely reactive, it doesn\u0027t really need to be part of a running agent and done via inter-process communication... Just run the required functions directly in the API Manager code.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":109,"context_line":"When a request for a new instance or failover is received the Build Manager"},{"line_number":110,"context_line":"will have responsibility for connecting the appropriate Neutron ports to the"},{"line_number":111,"context_line":"Amphora via the network driver and triggering a configuration push via the"},{"line_number":112,"context_line":"Amphora driver."},{"line_number":113,"context_line":""},{"line_number":114,"context_line":"The Amphora configured by the Build Manager may be an existing Amphora"},{"line_number":115,"context_line":"instance or a new Amphora from the spares pool.  This determination will"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_6b751df9","line":112,"in_reply_to":"5a890539_07008a01","updated":"2014-12-11 00:12:42.000000000","message":"See my comment above.  The wire up process, HA setups, and scalability configurations (thing many AMPs for one listener, any of which could fail to startup) could take a long time.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":114,"context_line":"The Amphora configured by the Build Manager may be an existing Amphora"},{"line_number":115,"context_line":"instance or a new Amphora from the spares pool.  This determination will"},{"line_number":116,"context_line":"be made based on the apolocation requirements of the load balancer and the"},{"line_number":117,"context_line":"load balancer count on the existing Amphora."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"The main flow of the Build Manager is described in the"},{"line_number":120,"context_line":"amphora-lifecycle-management specification as the Activate Amphora sequence."}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_67a86e8d","line":117,"updated":"2014-12-05 07:33:46.000000000","message":"Should also note that this can also be a brand new Amphora because the spare amphora pool size might be 0, in which case there is no spare amphora pool.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":114,"context_line":"The Amphora configured by the Build Manager may be an existing Amphora"},{"line_number":115,"context_line":"instance or a new Amphora from the spares pool.  This determination will"},{"line_number":116,"context_line":"be made based on the apolocation requirements of the load balancer and the"},{"line_number":117,"context_line":"load balancer count on the existing Amphora."},{"line_number":118,"context_line":""},{"line_number":119,"context_line":"The main flow of the Build Manager is described in the"},{"line_number":120,"context_line":"amphora-lifecycle-management specification as the Activate Amphora sequence."}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_6b4c3dc9","line":117,"in_reply_to":"5a890539_67a86e8d","updated":"2014-12-11 00:12:42.000000000","message":"Good catch, updated in the next patch set.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":124,"context_line":"The Certificate Manager responds to requests for security data stored in"},{"line_number":125,"context_line":"OpenStack Barbican from the Amphora Driver.  It will provide a short term"},{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_933be438","line":127,"updated":"2014-12-05 04:30:08.000000000","message":"Good idea! Cache should be in memory only. :D","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":124,"context_line":"The Certificate Manager responds to requests for security data stored in"},{"line_number":125,"context_line":"OpenStack Barbican from the Amphora Driver.  It will provide a short term"},{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_a72176b7","line":127,"in_reply_to":"5a890539_933be438","updated":"2014-12-05 08:00:24.000000000","message":"Hmm, alright -- I don\u0027t know what I think about this from a security perspective, but I guess it\u0027s acceptable. Though, if we ever get transport-key wrapping implemented between Amphorae and Barbican/other, caching would not be viable.\n\nNo cache invalidation because we\u0027re assuming Certs/Keys are immutable in whatever system we use. :)\nThat only becomes a problem when that assumption changes...","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":124,"context_line":"The Certificate Manager responds to requests for security data stored in"},{"line_number":125,"context_line":"OpenStack Barbican from the Amphora Driver.  It will provide a short term"},{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_07708a3f","line":127,"in_reply_to":"5a890539_933be438","updated":"2014-12-05 07:33:46.000000000","message":"oh no cache invalidation!","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":124,"context_line":"The Certificate Manager responds to requests for security data stored in"},{"line_number":125,"context_line":"OpenStack Barbican from the Amphora Driver.  It will provide a short term"},{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_ebe66daf","line":127,"in_reply_to":"5a890539_a72176b7","updated":"2014-12-11 00:12:42.000000000","message":"I am thinking about customers that start up hundreds of load balancer instances using tools like puppet (we have some).","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d017a55d65e0ca767603e286a12e90f53dc0af13","unresolved":false,"context_lines":[{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"The Health Manager is tasked with receiving health and heartbeat messages from"},{"line_number":132,"context_line":"the Octavia Controller Message Manager.  Health event information will be"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_f370a80a","line":129,"updated":"2014-12-05 04:30:08.000000000","message":"So, everything you\u0027ve said below about \"listeners\" actually applies to amphorae. Amphorae and listeners are not the same thing, eh-- and we don\u0027t necessarily want to trigger a failover on a failed listener. (We may simply want to try restarting it first, and then trigger a failover if it doesn\u0027t succeed after a certain number of attempts.)","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"The Health Manager is tasked with receiving health and heartbeat messages from"},{"line_number":132,"context_line":"the Octavia Controller Message Manager.  Health event information will be"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_2b7955c8","line":129,"in_reply_to":"5a890539_07ec0abb","updated":"2014-12-11 00:12:42.000000000","message":"Hmmm, Since we have separate processes, one per listener, we cannot assume that we can only monitor the Amphora at this level.\nMy assumption is restarting the haproxy process would be handled by the agent and would happen before the agent reports a listener issue to the Health Manager.  I am assuming that if the agent reports to the controller that the listener is failed it has exhausted the recovery options it has available and it needs the controller to fail over the listener.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":126,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":127,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":128,"context_line":""},{"line_number":129,"context_line":"**Health Manager**"},{"line_number":130,"context_line":""},{"line_number":131,"context_line":"The Health Manager is tasked with receiving health and heartbeat messages from"},{"line_number":132,"context_line":"the Octavia Controller Message Manager.  Health event information will be"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_07ec0abb","line":129,"in_reply_to":"5a890539_f370a80a","updated":"2014-12-05 07:33:46.000000000","message":"I agree with stephen, this is supposed to be monitoring Amphorae.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":133,"context_line":"stored in the highly available database.  Each listener will have an entry in a"},{"line_number":134,"context_line":"dictionary that holds it\u0027s last heartbeat timestamp.  The"},{"line_number":135,"context_line":"amphora-lifecycle-management specification details the health monitoring"},{"line_number":136,"context_line":"sequence."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The health monitor will have a separate thread that checks these timestamps on"},{"line_number":139,"context_line":"a configurable interval to see if the listener has not provided a heartbeat in"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_02179846","line":136,"updated":"2014-12-05 08:00:24.000000000","message":"Yes, as Stephen said, it\u0027s really per Amphora, not per listener, that we deal with heartbeats. So if Listener A and B are both on Amphora 1, they are both up so long as Amphora 1 is up (for some definition of up, since the heartbeat is just coming from the Amphora not from the individual HAProxy instances), though I suppose it would be possible to include some sort of status in the heartbeat message... Of course, now we\u0027re getting into details that might not belong here.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":133,"context_line":"stored in the highly available database.  Each listener will have an entry in a"},{"line_number":134,"context_line":"dictionary that holds it\u0027s last heartbeat timestamp.  The"},{"line_number":135,"context_line":"amphora-lifecycle-management specification details the health monitoring"},{"line_number":136,"context_line":"sequence."},{"line_number":137,"context_line":""},{"line_number":138,"context_line":"The health monitor will have a separate thread that checks these timestamps on"},{"line_number":139,"context_line":"a configurable interval to see if the listener has not provided a heartbeat in"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_0b3f39d7","line":136,"in_reply_to":"5a890539_02179846","updated":"2014-12-11 00:12:42.000000000","message":"We voted that each listener is a separate haproxy instance, so we cannot assume that Listener A is healthy if Listener B is....","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":144,"context_line":""},{"line_number":145,"context_line":"The Health Manager will have to be aware of the load balancer associated with"},{"line_number":146,"context_line":"the failed listener to decide if it needs to fail over additional listeners to"},{"line_number":147,"context_line":"migrate the failed listener to a new Amphora."},{"line_number":148,"context_line":""},{"line_number":149,"context_line":"**Housekeeping Manager**"},{"line_number":150,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_02b3184e","line":147,"updated":"2014-12-05 08:00:24.000000000","message":"So this is an interesting point -- until now I haven\u0027t been considering multiple Listeners per Amphora, which I suppose is required for shared VIPs. This complicates things a bit. We may need some additional outside monitoring (ping/connect/etc) to actually verify the status of our listeners, as I don\u0027t know if even inspecting the HAProxy service *on the Amphora* is going to tell us much, since it could be \"up\" but not actually functioning, or functioning and connectable locally but not externally.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"9974ccd3f570ce975faa64f0b0c117b195f8c84a","unresolved":false,"context_lines":[{"line_number":144,"context_line":""},{"line_number":145,"context_line":"The Health Manager will have to be aware of the load balancer associated with"},{"line_number":146,"context_line":"the failed listener to decide if it needs to fail over additional listeners to"},{"line_number":147,"context_line":"migrate the failed listener to a new Amphora."},{"line_number":148,"context_line":""},{"line_number":149,"context_line":"**Housekeeping Manager**"},{"line_number":150,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_ba97acea","line":147,"in_reply_to":"5a890539_02b3184e","updated":"2014-12-06 00:37:48.000000000","message":"+1","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":153,"context_line":"Housekeeping Manager will check the Octavia database to identify the required"},{"line_number":154,"context_line":"cleanup and maintenance actions required.  The amphora-lifecycle-management"},{"line_number":155,"context_line":"specification details the Create, Spare, and Delete amphora sequences the"},{"line_number":156,"context_line":"Housekeeping Manager will follow."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"The operator can specify a number of Amphora instances to be held in a spares"},{"line_number":159,"context_line":"pool.  Building Amphora instances can take a long time so the Housekeeping"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_a278842d","line":156,"updated":"2014-12-05 08:00:24.000000000","message":"See, you\u0027re already detailing all of these operations here -- I think probably just remove the Build Manager entirely, call this the Amphora Lifecycle Manager, and from the old Build Manager section merge Create/Delete into this and Update into the API Manager.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":153,"context_line":"Housekeeping Manager will check the Octavia database to identify the required"},{"line_number":154,"context_line":"cleanup and maintenance actions required.  The amphora-lifecycle-management"},{"line_number":155,"context_line":"specification details the Create, Spare, and Delete amphora sequences the"},{"line_number":156,"context_line":"Housekeeping Manager will follow."},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"The operator can specify a number of Amphora instances to be held in a spares"},{"line_number":159,"context_line":"pool.  Building Amphora instances can take a long time so the Housekeeping"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_f6073016","line":156,"in_reply_to":"5a890539_a278842d","updated":"2014-12-11 00:12:42.000000000","message":"See notes above about the differences between Housekeeping Manager and the component formerly known as \"Build Manager\"","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":10273,"name":"Adam Harwell","email":"flux.adam@gmail.com","username":"rm_you"},"change_message_id":"4f9ed4d078babaaf40e71633e6f74862bcce677d","unresolved":false,"context_lines":[{"line_number":160,"context_line":"Manager will spawn threads to manage the spare Amphora lifecycle."},{"line_number":161,"context_line":""},{"line_number":162,"context_line":"The Housekeeping Manager will interface with the compute driver, network driver,"},{"line_number":163,"context_line":"and the Certificate Manager to accomplish the create and delete actions."},{"line_number":164,"context_line":""},{"line_number":165,"context_line":"**Message Manager**"},{"line_number":166,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_228c942e","line":163,"updated":"2014-12-05 08:00:24.000000000","message":"If Update isn\u0027t part of Housekeeping/Lifecycle (which I don\u0027t think it should be), then it wouldn\u0027t  need to talk to the Cert manager.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":160,"context_line":"Manager will spawn threads to manage the spare Amphora lifecycle."},{"line_number":161,"context_line":""},{"line_number":162,"context_line":"The Housekeeping Manager will interface with the compute driver, network driver,"},{"line_number":163,"context_line":"and the Certificate Manager to accomplish the create and delete actions."},{"line_number":164,"context_line":""},{"line_number":165,"context_line":"**Message Manager**"},{"line_number":166,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_b6284887","line":163,"in_reply_to":"5a890539_228c942e","updated":"2014-12-11 00:12:42.000000000","message":"We don\u0027t need to \"de-register\" the use of a certificate after all of the listeners using the certificate are deleted?","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"abaaa91cf1e290d02fdda4d6ddceb26f496ad7a9","unresolved":false,"context_lines":[{"line_number":243,"context_line":""},{"line_number":244,"context_line":"References"},{"line_number":245,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5a890539_275d867d","line":246,"updated":"2014-12-05 07:33:46.000000000","message":"should reference the amphora lifecycle management review","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"776e90361648b156a13e29a32d7e776b33739250","unresolved":false,"context_lines":[{"line_number":243,"context_line":""},{"line_number":244,"context_line":"References"},{"line_number":245,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":246,"context_line":""},{"line_number":247,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"3a961159_1624bcad","line":246,"in_reply_to":"5a890539_275d867d","updated":"2014-12-11 00:12:42.000000000","message":"Good catch.  Included in the next patch set.","commit_id":"7536e0baca5b5bee7ec408a530d81abeb49cdc90"}],"specs/version0.5/controller.dot":[{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":1,"context_line":"﻿digraph G {"},{"line_number":2,"context_line":""},{"line_number":3,"context_line":""},{"line_number":4,"context_line":"   subgraph cluster0 {"}],"source_content_type":"application/msword","patch_set":6,"id":"3a961159_16ae7c98","line":1,"updated":"2014-12-18 23:29:53.000000000","message":"I\u0027m really impressed with how well you were able to get GraphViz to do what you want (it\u0027s not trivial...) Anyway, this doesn\u0027t seem to want to render for me unless there\u0027s a blank line or comment at the top. So could you put the standard copyright statement in a comment at the top of this file?  GraphViz does C-style comments, so:  /* Comment */","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"}],"specs/version0.5/controller.rst":[{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":36,"context_line":"* Respecting colocation / apolocation / flavor requirements of the Amphora"},{"line_number":37,"context_line":"* Processing statistical data from the Amphora including communicating with"},{"line_number":38,"context_line":"  metering services, such as Ceilometer"},{"line_number":39,"context_line":"  (https://blueprints.launchpad.net/ceilometer/+spec/ceilometer-meter-lbaas)"},{"line_number":40,"context_line":"* Responding to API requests sent by the API processes"},{"line_number":41,"context_line":"* Proxy Amphora data to other OpenStack services such as Swift for log file"},{"line_number":42,"context_line":"  archival"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_51a2ce99","line":39,"updated":"2014-12-11 03:19:14.000000000","message":"Having looked at this blueprint and the specs and code associated with them, we\u0027ll likely have to adjust things on the ceilometer side of things, as it looks like they\u0027re using LBaaS v1 object model concepts, which don\u0027t really apply here. (We\u0027re going to care about statistics corresponding to a given listener, not to a pool, for example.)","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":71,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":72,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":73,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":74,"context_line":"a given controller is primarily for distributing monitoring and network load."},{"line_number":75,"context_line":""},{"line_number":76,"context_line":"**Amphora Driver**"},{"line_number":77,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_f1249a61","line":74,"updated":"2014-12-11 03:19:14.000000000","message":"Continuing discussion from PS1: I shall come to the discussion prepared to talk about this design and the couple of bottlenecks I see in it as proposed. :)","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":115,"context_line":"The Amphora configured by the Deploy Manager may be an existing Amphora"},{"line_number":116,"context_line":"instance, a new Amphora from the spares pool, or a newly created Amphora."},{"line_number":117,"context_line":"This determination will be made based on the apolocation requirements of"},{"line_number":118,"context_line":"the load balancer and the load balancer count on the existing Amphora."},{"line_number":119,"context_line":""},{"line_number":120,"context_line":"The main flow of the Deploy Manager is described in the"},{"line_number":121,"context_line":"amphora-lifecycle-management specification as the Activate Amphora sequence."}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_2648a786","line":118,"updated":"2014-12-11 03:19:14.000000000","message":"Continuing discussion from PS1: From this document I gather both the Housekeeping manager and Deploy manager can spin up new amphorae. The latter puts these into the spares pool, while the former simply activates them immediately once they\u0027re up.\n\nDoes it therefore make sense to either delegate these entirely to the Housekeeping manager, or create a new manager / process which can be spawned from either the Deploy or Housekeeping manager for spinning up new amphorae (and keeps track of the number of amphorae currently in the process of being spun up so we have a check where we can throttle this if there\u0027s something significantly wrong with the cloud)?  (Also keep in mind that there\u0027s quite a bit that needs to happen when a new amphora is created.)","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":127,"context_line":"(1 minute) cache of the security contents to facilitate the efficient"},{"line_number":128,"context_line":"startup of a large number of listeners sharing security content."},{"line_number":129,"context_line":""},{"line_number":130,"context_line":"**Health Manager**"},{"line_number":131,"context_line":""},{"line_number":132,"context_line":"The Health Manager is tasked with receiving health and heartbeat messages from"},{"line_number":133,"context_line":"the Octavia Controller Message Manager.  Health event information will be"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_f9072016","line":130,"updated":"2014-12-11 03:19:14.000000000","message":"Continuing discussion from PS1:\n\nIf we\u0027re only monitoring listeners from the controller layer, how do you intend to monitor the health of an amphora if it doesn\u0027t have any listeners on it yet (ie. it\u0027s in the spares pool)?\n\nFurthermore, yes, having one listener fail on a given amphora repeatedly means that the whole loadbalancer needs to be migrated elsewhere. The logic behind this can be described in a pretty simple flow-chart (which we should white-board out at the hack-a-thon).\n\nInterestingly enough: Depending on the capabilities of upstream routing (ie. is it layer-4 aware?) we may be able to have two listeners on the same load balancer (and VIP) serviced by separate amphorae at some point. Keeping the listeners separate allows for this possibility too-- and if we can do that, a single failed listener doesn\u0027t necessarily mean the whole loadbalancer needs to fail over.\n\nFor now, it\u0027s probably safe to assume that if the amphora is reachable and its listeners are running, the listener is probably healthy (99.99% of the time).","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":145,"context_line":""},{"line_number":146,"context_line":"The Health Manager will have to be aware of the load balancer associated with"},{"line_number":147,"context_line":"the failed listener to decide if it needs to fail over additional listeners to"},{"line_number":148,"context_line":"migrate the failed listener to a new Amphora."},{"line_number":149,"context_line":""},{"line_number":150,"context_line":"**Housekeeping Manager**"},{"line_number":151,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_39b1a847","line":148,"updated":"2014-12-11 03:19:14.000000000","message":"Continuing discussion from PS1:\n\nThere are a whole bunch of different ways listeners can fail that are only detectable by a specific kind of probe. Rather than trying to guess what the likely failures are, let\u0027s avoid trying to solve a problem that may not exist, and just go with something like this:\n\n* If the amphora is reachable from the controler.\n* AND If the listeners are running\n-\u003e Then the listeners are probably healthy.\n\n(which will probably be true 99.99% of the time.)  If we start to see other kinds of failures here then we should figure out ways to reliably detect and correct them.","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3c88517583d0cac00716f159d97a49d68af7f5e2","unresolved":false,"context_lines":[{"line_number":145,"context_line":""},{"line_number":146,"context_line":"The Health Manager will have to be aware of the load balancer associated with"},{"line_number":147,"context_line":"the failed listener to decide if it needs to fail over additional listeners to"},{"line_number":148,"context_line":"migrate the failed listener to a new Amphora."},{"line_number":149,"context_line":""},{"line_number":150,"context_line":"**Housekeeping Manager**"},{"line_number":151,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_49976dec","line":148,"in_reply_to":"3a961159_39b1a847","updated":"2014-12-11 17:30:47.000000000","message":"well, the spec is vague enough that \"failing over a listener\" can mean restarting that listener, bringing in a new amphora (if this was the only listener), etc. -- maybe we should use as wording \"repair and eventually fail over\" --","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"cdc38b7db624f3dafdbbd70f43a9e8cfd969a52f","unresolved":false,"context_lines":[{"line_number":161,"context_line":"Manager will spawn threads to manage the spare Amphora lifecycle."},{"line_number":162,"context_line":""},{"line_number":163,"context_line":"The Housekeeping Manager will interface with the compute driver, network driver,"},{"line_number":164,"context_line":"and the Certificate Manager to accomplish the create and delete actions."},{"line_number":165,"context_line":""},{"line_number":166,"context_line":"**Message Manager**"},{"line_number":167,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_39dfe86a","line":164,"updated":"2014-12-11 03:19:14.000000000","message":"Continuing discussion from PS1:\n\n\u0027Delete\u0027 operation here deletes an amphora, correct? I would think that the amphora driver would be invoked first to delete any outstanding listeners (and would be the logical thing to clean up barbican registrations)...  In my mind, by the time the housekeeping manager is called to delete an amphora, that amphora should essentially be empty or non-functional.","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":10850,"name":"German Eichberger","email":"german.eichberger@gmail.com","username":"german"},"change_message_id":"3c88517583d0cac00716f159d97a49d68af7f5e2","unresolved":false,"context_lines":[{"line_number":161,"context_line":"Manager will spawn threads to manage the spare Amphora lifecycle."},{"line_number":162,"context_line":""},{"line_number":163,"context_line":"The Housekeeping Manager will interface with the compute driver, network driver,"},{"line_number":164,"context_line":"and the Certificate Manager to accomplish the create and delete actions."},{"line_number":165,"context_line":""},{"line_number":166,"context_line":"**Message Manager**"},{"line_number":167,"context_line":""}],"source_content_type":"text/x-rst","patch_set":2,"id":"3a961159_494ccd9e","line":164,"in_reply_to":"3a961159_39dfe86a","updated":"2014-12-11 17:30:47.000000000","message":"The background is that if you start deleting more than a couple of nova vms at the same time nova might explode  so having that allows for rate limiting the deletes. Again, this needs some testing but not having that option and discovering that nova/docker/etc. explodes would be back for an operator...","commit_id":"f708b09ea16a077182097a628070a5738d7ebf30"},{"author":{"_account_id":8179,"name":"Jamie McCarthy","email":"jamie@mccarthy.vg","username":"jamiemccarthy"},"change_message_id":"6d9c3cc9ec0082a2df4055cd53b1b04bb058eb04","unresolved":false,"context_lines":[{"line_number":65,"context_line":"information about the Amphora, load balancers, and listeners via the database."},{"line_number":66,"context_line":"It is expected that the Octavia controller will directly interact with the"},{"line_number":67,"context_line":"database but the Amphora driver and Amphorae will not directly access the"},{"line_number":68,"context_line":"database."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"}],"source_content_type":"text/x-rst","patch_set":4,"id":"3a961159_f354cdcb","line":68,"updated":"2014-12-17 23:11:49.000000000","message":"I think we decided at this week\u0027s hackathon that the Message Manager is gone, in which case the Message Handler (part of the Amphora driver) would indeed directly access the database. If that\u0027s correct, then the above sentence should be changed.","commit_id":"eb7dfe6878e055e6ff909874f9bc536a8831433b"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":49,"context_line":"* Amphora Driver"},{"line_number":50,"context_line":"* API Manager"},{"line_number":51,"context_line":"* Deploy Manager"},{"line_number":52,"context_line":"* Certificate Manager"},{"line_number":53,"context_line":"* Health Manager"},{"line_number":54,"context_line":"* Housekeeping Manager"},{"line_number":55,"context_line":"* Services Proxy"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_702c28ad","line":52,"updated":"2014-12-18 23:29:53.000000000","message":"I think we decided there\u0027s actually going to be a library for this.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":64,"context_line":"Octavia controller.  Octavia controller processes will share state and"},{"line_number":65,"context_line":"information about the Amphora, load balancers, and listeners via the database."},{"line_number":66,"context_line":"It is expected that the Octavia controller and Amphora driver will directly"},{"line_number":67,"context_line":"interact with the database but the Amphorae will not directly access the"},{"line_number":68,"context_line":"database."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_cb7a9187","line":67,"updated":"2014-12-18 23:29:53.000000000","message":"Let\u0027s use stronger language here: The amphorae should never directly access the database. (Too much of a security risk to allow this.)","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":67,"context_line":"interact with the database but the Amphorae will not directly access the"},{"line_number":68,"context_line":"database."},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"By using a highly available database, Octavia controllers, of a certain flavor,"},{"line_number":71,"context_line":"can service API messaging queues and Amphora events regardless of if the"},{"line_number":72,"context_line":"servicing controller \"owns\" the affected Amphora.  Ownership of an Amphora by"},{"line_number":73,"context_line":"a given controller is primarily for distributing monitoring and network load."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_8be1692c","line":70,"updated":"2014-12-18 23:29:53.000000000","message":"The wording of this paragraph is pretty awkward. Let me suggest the following:\n\nBy using a highly available database, Octavia controllers themselves do not directly keep any stateful information on amphorae. Because of this, amphorae are not assigned to any specific controller (except for the duration of a single heartbeat). Any controller is able to service monitoring, heartbeat, API, and other requests coming to or from amphorae.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":82,"context_line":""},{"line_number":83,"context_line":"The API Manager is event driven and tasked with servicing requests from the"},{"line_number":84,"context_line":"API components via an Oslo messaging queue.  It is also the primary lifecycle"},{"line_number":85,"context_line":"management component for Amphora.  Most API requests can be serviced by"},{"line_number":86,"context_line":"querying the database for the current state of load balancers and listeners."},{"line_number":87,"context_line":""},{"line_number":88,"context_line":"Requests for a new load balancer or listener will cause the API Manager to"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_ab05457b","line":85,"updated":"2014-12-18 23:29:53.000000000","message":"Nit: s/Most/Many/","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":92,"context_line":""},{"line_number":93,"context_line":"API requests may also cause the API Manager to push updated configurations to"},{"line_number":94,"context_line":"the Octavia database and notify the driver to push an updated configuration"},{"line_number":95,"context_line":"to the listener."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Delete requests for load balancers or listeners will call the Amphora driver"},{"line_number":98,"context_line":"wrap up method to gracefully shutdown the processes, ship logs, ship"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_0b0f3989","line":95,"updated":"2014-12-18 23:29:53.000000000","message":"Also note: A worker will be dispatched to supervise the process of pushing out the new configuration.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":102,"context_line":""},{"line_number":103,"context_line":"**Deploy Worker**"},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"The Deploy Worker is a spawned process from the API Manager or the Health"},{"line_number":106,"context_line":"Manager.  It interfaces with the compute driver (in some deployment scenarios),"},{"line_number":107,"context_line":"network driver, and Amphora driver to activate Amphora instances,"},{"line_number":108,"context_line":"load balancers, and listeners."}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_6b1d1dbe","line":105,"updated":"2014-12-18 23:29:53.000000000","message":"or thread?","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":116,"context_line":"The Amphora configured by the Deploy Worker may be an existing Amphora"},{"line_number":117,"context_line":"instance, a new Amphora from the spares pool, or a newly created Amphora."},{"line_number":118,"context_line":"This determination will be made based on the apolocation requirements of"},{"line_number":119,"context_line":"the load balancer and the load balancer count on the existing Amphora."},{"line_number":120,"context_line":""},{"line_number":121,"context_line":"The Deploy Worker will be resposible for passing in the required metadata"},{"line_number":122,"context_line":"via config drive when deploying an Amphora.  This metadata will include:"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_8bdfa9d5","line":119,"updated":"2014-12-18 23:29:53.000000000","message":"And the availability of ready spare amphorae in the spares pool.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":155,"context_line":"The Housekeeping Manager will manage the spare Amphora pool and the teardown"},{"line_number":156,"context_line":"of Amphora that are no longer needed.  On a configurable interval the"},{"line_number":157,"context_line":"Housekeeping Manager will check the Octavia database to identify the required"},{"line_number":158,"context_line":"cleanup and maintenance actions required.  The amphora-lifecycle-management"},{"line_number":159,"context_line":"specification details the Create, Spare, and Delete amphora sequences the"},{"line_number":160,"context_line":"Housekeeping Manager will follow."},{"line_number":161,"context_line":""}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_3697f891","line":158,"updated":"2014-12-18 23:29:53.000000000","message":"Nit: Kill one of the \u0027required\u0027 in that sentence.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"bb62a9f03d4ce37f779f4b391b3021eb46179db7","unresolved":false,"context_lines":[{"line_number":161,"context_line":""},{"line_number":162,"context_line":"The operator can specify a number of Amphora instances to be held in a spares"},{"line_number":163,"context_line":"pool.  Building Amphora instances can take a long time so the Housekeeping"},{"line_number":164,"context_line":"Manager will spawn threads to manage the spare Amphora lifecycle."},{"line_number":165,"context_line":""},{"line_number":166,"context_line":"The Housekeeping Manager will interface with the compute driver,"},{"line_number":167,"context_line":"network driver, and the Certificate Manager to accomplish the create"}],"source_content_type":"text/x-rst","patch_set":6,"id":"3a961159_76e4203c","line":164,"updated":"2014-12-18 23:29:53.000000000","message":"Nit: to manage the number of amphorae in the spares pool.","commit_id":"9fa87696dbab19c541b651dde87cc572f6494bcf"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"573e85ad5bb48ed074208233cefd9e187046de83","unresolved":false,"context_lines":[{"line_number":46,"context_line":""},{"line_number":47,"context_line":"The Octavia controller will consist of the following components:"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"* Amphora Driver"},{"line_number":50,"context_line":"* API Manager"},{"line_number":51,"context_line":"* Deploy Manager"},{"line_number":52,"context_line":"* Certificate Library"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_7293455b","line":49,"updated":"2014-12-19 06:37:09.000000000","message":"If you\u0027re including the Amphora driver, might as well include compute driver and network driver.  I\u0027d suggest not including any of these drivers in this list, as they are more implementation details.","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d8d7bc3e646262c1eb30981a7bfac7c0706d7687","unresolved":false,"context_lines":[{"line_number":46,"context_line":""},{"line_number":47,"context_line":"The Octavia controller will consist of the following components:"},{"line_number":48,"context_line":""},{"line_number":49,"context_line":"* Amphora Driver"},{"line_number":50,"context_line":"* API Manager"},{"line_number":51,"context_line":"* Deploy Manager"},{"line_number":52,"context_line":"* Certificate Library"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_800f6a38","line":49,"in_reply_to":"3a961159_7293455b","updated":"2014-12-19 18:55:12.000000000","message":"Agreed. I think the drivers probably should be mentioned since they ultimately become part of the controller (talking to a compute service, networking service, and amphora(e) respectively.  But I agree that you should mention all of them or nothing.","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"573e85ad5bb48ed074208233cefd9e187046de83","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"* Amphora Driver"},{"line_number":50,"context_line":"* API Manager"},{"line_number":51,"context_line":"* Deploy Manager"},{"line_number":52,"context_line":"* Certificate Library"},{"line_number":53,"context_line":"* Health Manager"},{"line_number":54,"context_line":"* Housekeeping Manager"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_f2a735bb","line":51,"updated":"2014-12-19 06:37:09.000000000","message":"nit: Deploy Worker","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d8d7bc3e646262c1eb30981a7bfac7c0706d7687","unresolved":false,"context_lines":[{"line_number":48,"context_line":""},{"line_number":49,"context_line":"* Amphora Driver"},{"line_number":50,"context_line":"* API Manager"},{"line_number":51,"context_line":"* Deploy Manager"},{"line_number":52,"context_line":"* Certificate Library"},{"line_number":53,"context_line":"* Health Manager"},{"line_number":54,"context_line":"* Housekeeping Manager"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_a014664d","line":51,"in_reply_to":"3a961159_f2a735bb","updated":"2014-12-19 18:55:12.000000000","message":"+1","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"573e85ad5bb48ed074208233cefd9e187046de83","unresolved":false,"context_lines":[{"line_number":58,"context_line":""},{"line_number":59,"context_line":"These components should be implemented as independent processes that"},{"line_number":60,"context_line":"communicate with each other via zeromq unless otherwise noted.  This provides"},{"line_number":61,"context_line":"a level of autonomy to these functions."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The highly available database will provide the persistent \"brain\" for the"},{"line_number":64,"context_line":"Octavia controller.  Octavia controller processes will share state and"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_1297613b","line":61,"updated":"2014-12-19 06:37:09.000000000","message":"I don\u0027t believe this stands still.  I don\u0027t think these workers are communicating with each other through any queue (other than API Manager receiving messages over the queue).  Aren\u0027t they strictly making decisions based on what is the database? for now at least.","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":11685,"name":"Stephen Balukoff","email":"stephen@balukoff.com","username":"sbalukoff"},"change_message_id":"d8d7bc3e646262c1eb30981a7bfac7c0706d7687","unresolved":false,"context_lines":[{"line_number":58,"context_line":""},{"line_number":59,"context_line":"These components should be implemented as independent processes that"},{"line_number":60,"context_line":"communicate with each other via zeromq unless otherwise noted.  This provides"},{"line_number":61,"context_line":"a level of autonomy to these functions."},{"line_number":62,"context_line":""},{"line_number":63,"context_line":"The highly available database will provide the persistent \"brain\" for the"},{"line_number":64,"context_line":"Octavia controller.  Octavia controller processes will share state and"}],"source_content_type":"text/x-rst","patch_set":8,"id":"3a961159_63052810","line":61,"in_reply_to":"3a961159_1297613b","updated":"2014-12-19 18:55:12.000000000","message":"I... think you\u0027re correct. And in any case, any kind of tasks we\u0027re going to be calling will be using a worker thread, which is likely to call on driver methods, but probably won\u0027t ever be communicating directly with any of the managers.","commit_id":"65897533d1a72bec9fe59dc736990e05fed9021f"},{"author":{"_account_id":10750,"name":"Jorge Miramontes","email":"jorge.miramontes@rackspace.com","username":"jorgem"},"change_message_id":"3fe7ff8fbf67ff865f8324e234e698360821de69","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_c046a3bf","line":104,"updated":"2015-01-07 22:38:29.000000000","message":"I know this is merged already, but I\u0027m a bit confused on the consensus for the API Manager. For example, is a deploy worker spawned for each create/update/delete request? If this is the case, then it doesn\u0027t make much sense to have the API Manager access the database. The deploy worker will handle that. I see the API Manager really as a request delegator. If it only spawns a deploy worker for create requests then why does it not for other requests?","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"},{"author":{"_account_id":10750,"name":"Jorge Miramontes","email":"jorge.miramontes@rackspace.com","username":"jorgem"},"change_message_id":"6f88c94b6f297f8a33abf09cb57218598126a3bd","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_bcc15b31","line":104,"in_reply_to":"3a961159_911ac816","updated":"2015-01-12 18:10:00.000000000","message":"Sweet. Thanks!","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"e972c8a50d2864576382652865651dcfe8e8dd7e","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_911ac816","line":104,"in_reply_to":"3a961159_95e59fa5","updated":"2015-01-12 17:31:37.000000000","message":"My intention is that the API Manager would not need a taskflow flow, but the workers will be implemented as flows.\nI will be submitting a patch that clarifies this section.","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"},{"author":{"_account_id":11628,"name":"Michael Johnson","email":"johnsomor@gmail.com","username":"johnsom"},"change_message_id":"d00dcd64a156771ee5b1f90477c978051ad2be4b","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_ec6ab9eb","line":104,"in_reply_to":"3a961159_c046a3bf","updated":"2015-01-08 00:51:20.000000000","message":"I see the issue.  The section on the API manager servicing database queries should have been removed when we decided the API components outside the controller will have access to the Octavia database.\nThis section should also be updated with the new queue concepts that the API components outside of the controller will update the database and basically notify this component via the oslo messaging to spawn a deploy worker.","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"},{"author":{"_account_id":6951,"name":"Brandon Logan","email":"brandon.logan@rackspace.com","username":"brandon-logan"},"change_message_id":"801ddd7096bea1b64112f0a351189f79806d070f","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_80a2d43c","line":104,"in_reply_to":"3a961159_ec6ab9eb","updated":"2015-01-08 04:09:48.000000000","message":"I think Jorge just wants to know if the API Manager will only spawn a deploy worker for creates, and do something else for updates and deletes, or will it spawn a deploy worker for all CUD operations.","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"},{"author":{"_account_id":10750,"name":"Jorge Miramontes","email":"jorge.miramontes@rackspace.com","username":"jorgem"},"change_message_id":"b8895fd37f7cd352750580e9290d9501e4fc8d45","unresolved":false,"context_lines":[{"line_number":101,"context_line":"statistics, un-plug network resources, and mark pending delete status in the"},{"line_number":102,"context_line":"database for the Housekeeping Manager to cleanup the Amphora should it no"},{"line_number":103,"context_line":"longer be needed."},{"line_number":104,"context_line":""},{"line_number":105,"context_line":"**Compute Driver**"},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"The Compute Driver abstracts the implementation of instantiating the virtual"}],"source_content_type":"text/x-rst","patch_set":9,"id":"3a961159_95e59fa5","line":104,"in_reply_to":"3a961159_ec6ab9eb","updated":"2015-01-08 18:00:50.000000000","message":"Okay good to hear that this wasn\u0027t the case. Here\u0027s how I understand the API Manager to work:\n\n1) API Manager consumes create/update/delete message from the queue\n2) API Manager spawns a deploy worker with the message. I\u0027m assuming the message format will include the load balancer id and the action to take (ex. update pool member).\n\nTalking with James Arendt, he was trying to figure out if taskflow was going to be incorporated into the API Manager. I\u0027m interested to see where it would fit in because spawning deploy workers seems like the only responsibility of the API Manager. The workers handle the heavy lifting. Thoughts?","commit_id":"6ab85b6db9759b608af1e697b8dd2698e8f4c2dc"}]}
