)]}'
{"specs/newton/approved/instance-availability-profile.rst":[{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"6925d587fa6df9ebf11c302bb30244d17bb2fd63","unresolved":false,"context_lines":[{"line_number":77,"context_line":""},{"line_number":78,"context_line":"Initially we will define the following availability profile items:"},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"importance\u003d\u003c\u0027cattle\u0027, \u0027pet\u0027, \u0027panda\u0027\u003e"},{"line_number":81,"context_line":"down_time_max\u003d\u003cmax down time in seconds\u003e"},{"line_number":82,"context_line":"cpu_throttling_max\u003d\u003cmax cpu throttling percentage\u003e"},{"line_number":83,"context_line":"tolerate_suspend\u003d\u003cboolean\u003e"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a122d0e_e9742137","line":80,"range":{"start_line":80,"start_character":35,"end_line":80,"end_character":37},"updated":"2016-04-19 11:55:24.000000000","message":"As mentioned before I think these kind of super abstract terms are not something we should, as they have no clear meaning whatsoever.","commit_id":"9ac4bf4f574c3478651d2a3d13be3b076a4805ff"},{"author":{"_account_id":6608,"name":"Paul Carlton","email":"paul.carlton2@hpe.com","username":"paul-carlton2"},"change_message_id":"da2f46987f8a6664f1cd23d33c9579ee33c4428f","unresolved":false,"context_lines":[{"line_number":77,"context_line":""},{"line_number":78,"context_line":"Initially we will define the following availability profile items:"},{"line_number":79,"context_line":""},{"line_number":80,"context_line":"importance\u003d\u003c\u0027cattle\u0027, \u0027pet\u0027, \u0027panda\u0027\u003e"},{"line_number":81,"context_line":"down_time_max\u003d\u003cmax down time in seconds\u003e"},{"line_number":82,"context_line":"cpu_throttling_max\u003d\u003cmax cpu throttling percentage\u003e"},{"line_number":83,"context_line":"tolerate_suspend\u003d\u003cboolean\u003e"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a122d0e_117a1288","line":80,"range":{"start_line":80,"start_character":35,"end_line":80,"end_character":37},"in_reply_to":"1a122d0e_e9742137","updated":"2016-04-20 16:44:25.000000000","message":"Maybe we should drop the importance concept and instead focus on specific setting like the ones below.","commit_id":"9ac4bf4f574c3478651d2a3d13be3b076a4805ff"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"6925d587fa6df9ebf11c302bb30244d17bb2fd63","unresolved":false,"context_lines":[{"line_number":193,"context_line":"        shared TINYINT(1) DEFAULT 0,"},{"line_number":194,"context_line":"        user_id VARCHAR(255) NOT NULL,"},{"line_number":195,"context_line":"        project_id VARCHAR(255) NOT NULL,"},{"line_number":196,"context_line":"        INDEX tenant (project_id));"},{"line_number":197,"context_line":""},{"line_number":198,"context_line":"Add new table called instance_profile_items:"},{"line_number":199,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a122d0e_29df390e","line":196,"updated":"2016-04-19 11:55:24.000000000","message":"I\u0027m not convinced of the need for creating new APIs \u0026 tables for this. It seems the same goal could be achieve by simply using flavour extra specs to define SLA characteristics for the flavour.","commit_id":"9ac4bf4f574c3478651d2a3d13be3b076a4805ff"},{"author":{"_account_id":6608,"name":"Paul Carlton","email":"paul.carlton2@hpe.com","username":"paul-carlton2"},"change_message_id":"da2f46987f8a6664f1cd23d33c9579ee33c4428f","unresolved":false,"context_lines":[{"line_number":193,"context_line":"        shared TINYINT(1) DEFAULT 0,"},{"line_number":194,"context_line":"        user_id VARCHAR(255) NOT NULL,"},{"line_number":195,"context_line":"        project_id VARCHAR(255) NOT NULL,"},{"line_number":196,"context_line":"        INDEX tenant (project_id));"},{"line_number":197,"context_line":""},{"line_number":198,"context_line":"Add new table called instance_profile_items:"},{"line_number":199,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a122d0e_461b3281","line":196,"in_reply_to":"1a122d0e_29df390e","updated":"2016-04-20 16:44:25.000000000","message":"I considered that approach but rejected it because it could lead to an explosion of flavors.  I also considered using scheduler hints but was concerned this would be seen as a misuse and like flavor extra specs once the request_spec object is created as far as I\u0027m aware there is no API to alter it. This would prevent the instance owner from altering the availability settings during the life time of the instance which I consider a reasonable use case.","commit_id":"9ac4bf4f574c3478651d2a3d13be3b076a4805ff"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"8a25c6dfbd24341326a0e0c3bdefa24af7340b55","unresolved":false,"context_lines":[{"line_number":193,"context_line":"        shared TINYINT(1) DEFAULT 0,"},{"line_number":194,"context_line":"        user_id VARCHAR(255) NOT NULL,"},{"line_number":195,"context_line":"        project_id VARCHAR(255) NOT NULL,"},{"line_number":196,"context_line":"        INDEX tenant (project_id));"},{"line_number":197,"context_line":""},{"line_number":198,"context_line":"Add new table called instance_profile_items:"},{"line_number":199,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a122d0e_cc2e418b","line":196,"in_reply_to":"1a122d0e_461b3281","updated":"2016-04-20 16:50:13.000000000","message":"The \u0027explosion of flavours\u0027 concern is not a new one - it is raised every time add support for a new extra spec. There are ideas to deal with the flavour \"explosion\" problem in a general manner though which I think would be preferrable to your suggestion here which is just re-inventing flavours by another name. IOW, I\u0027d suggest this spec just do it with flavours and leave the problem of solving the \"explosion\" problem to another spec","commit_id":"9ac4bf4f574c3478651d2a3d13be3b076a4805ff"}],"specs/newton/approved/instance-importance.rst":[{"author":{"_account_id":12898,"name":"Tony Breeds","email":"tony@bakeyournoodle.com","username":"tonyb"},"change_message_id":"68a129d7c23ed17b6f7bd7521f8e9a2af7715e96","unresolved":false,"context_lines":[{"line_number":9,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Blueprint:"},{"line_number":12,"context_line":"https://blueprints.launchpad.net/nova/+spec/instance-importance"},{"line_number":13,"context_line":""},{"line_number":14,"context_line":"At present there is no formal way to record how important an instance"},{"line_number":15,"context_line":"is to the owner. This makes it difficult for operations and support"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a18252c_2ebd76b7","line":12,"updated":"2016-04-14 07:01:42.000000000","message":"Doesn\u0027t exist.","commit_id":"ead3c4e45b89ab13b2d153653f9cb0b455454229"},{"author":{"_account_id":6062,"name":"jichenjc","email":"jichenjc@cn.ibm.com","username":"jichenjc"},"change_message_id":"c1c22920807df9d9e5508c2d105532976353a808","unresolved":false,"context_lines":[{"line_number":19,"context_line":"Problem description"},{"line_number":20,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":21,"context_line":""},{"line_number":22,"context_line":"The operator may need to take some action that will impact an instance."},{"line_number":23,"context_line":"However they currently have no way of telling how important the"},{"line_number":24,"context_line":"instance is to the owner. It may be part of a distributed cloud aware"},{"line_number":25,"context_line":"application that will start a replacement automatically without any"}],"source_content_type":"text/x-rst","patch_set":3,"id":"5a18252c_30682387","line":22,"range":{"start_line":22,"start_character":30,"end_line":22,"end_character":34},"updated":"2016-04-14 12:17:18.000000000","message":"https://github.com/openstack/nova/blob/master/nova/api/openstack/compute/server_tags.py\nthose server tags might be a more generic way?","commit_id":"76c16f1e180043ee604046f69b210fde97b9aee4"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"a9011a490c6f19a501e443efd3da014b85915d36","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Pets are instance which if stopped cause an application outage."},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"Pandas are high value instances which should not be interrupted unless"},{"line_number":62,"context_line":"absolutely necessary."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Add an instance boot parameter to allow the instance creator to specify"},{"line_number":65,"context_line":"the importance of the instance. The \u0027importance\u0027 setting will default"}],"source_content_type":"text/x-rst","patch_set":3,"id":"5a18252c_82857b08","line":62,"updated":"2016-04-14 08:38:22.000000000","message":"I think this \"importance\" concept is too vague \u0026 opaque to be practically useful in practice. Consider live migration, where if we want to guarantee completion we have a choice of setting a very large downtime value, using post-copy which has a small risk of failure preventing ability to rollback, or throttling the CPUs by a potentially large percentage.  All of these choices have a negative impact on the guest but if we have to choose between those 3 options, it is pretty much as clear as mud which would be most suitable for pets \u0026 pandas. Some pets may be fine with their CPUs being throttled, but most not have a large downtime value, while other pets would have outage if their CPU was thottled too much but would be fine accepting the tiny risk of post-copy failure, while other pets would not accept the risk of post-copy failure but would be fine with large downtime value.","commit_id":"76c16f1e180043ee604046f69b210fde97b9aee4"},{"author":{"_account_id":1779,"name":"Daniel Berrange","email":"berrange@redhat.com","username":"berrange"},"change_message_id":"0e6faf3b4e09c79e3970b3fb372fbe7d3d12ffff","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Pets are instance which if stopped cause an application outage."},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"Pandas are high value instances which should not be interrupted unless"},{"line_number":62,"context_line":"absolutely necessary."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Add an instance boot parameter to allow the instance creator to specify"},{"line_number":65,"context_line":"the importance of the instance. The \u0027importance\u0027 setting will default"}],"source_content_type":"text/x-rst","patch_set":3,"id":"5a18252c_538fb629","line":62,"in_reply_to":"5a18252c_13809f8a","updated":"2016-04-14 09:41:32.000000000","message":"Yep, I think some way to characterize the performance \u0026 reliability SLAs would be a more suitable direction, and could allow us to extend it to other features. For example instead of letting people directly set CPU, disk i/o \u0026 net i/o tuning parameters against the VM which are inherantly virt driver specific, a performance SLA could enable us to abstract the concept. This does have the potential to become quite a large piece of work to analyse the broader requirements.","commit_id":"76c16f1e180043ee604046f69b210fde97b9aee4"},{"author":{"_account_id":6608,"name":"Paul Carlton","email":"paul.carlton2@hpe.com","username":"paul-carlton2"},"change_message_id":"02e486bbd487f910a4111a5c3526047605596f7b","unresolved":false,"context_lines":[{"line_number":59,"context_line":"Pets are instance which if stopped cause an application outage."},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"Pandas are high value instances which should not be interrupted unless"},{"line_number":62,"context_line":"absolutely necessary."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Add an instance boot parameter to allow the instance creator to specify"},{"line_number":65,"context_line":"the importance of the instance. The \u0027importance\u0027 setting will default"}],"source_content_type":"text/x-rst","patch_set":3,"id":"5a18252c_13809f8a","line":62,"in_reply_to":"5a18252c_82857b08","updated":"2016-04-14 09:32:51.000000000","message":"I get your point but was trying to keep this simple and not expose libvirt specific live migration features.  I am working on a separate specification to automate the use of features like post-copy and auto-converge. As discussed in the past we don\u0027t want to expose these features in nova but have live migration work out which one to use.\n\nI\u0027ve separated this specification form the live migration automation specification to avoid making the \u0027importance\u0027 enhancement specific to live migration.  However maybe we need to allow the instance owner to\nspecify an instance availability profile that describes the acceptable maximum for down time or cpu throttling etc and prioritizes these, i.e.  specify that an instance outage is more acceptable than reduction in CPU power or vice versa.","commit_id":"76c16f1e180043ee604046f69b210fde97b9aee4"}]}