)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":14511,"name":"Maciej Szankin","email":"maciej+review@szankin.pl","username":"mszankin"},"change_message_id":"1d94419470abb0ed2a444f7b67f7f086a4979f69","unresolved":false,"context_lines":[{"line_number":4,"context_line":"Commit:     Maciej Szankin \u003cmaciej.szankin@intel.com\u003e"},{"line_number":5,"context_line":"CommitDate: 2017-04-02 23:35:50 -0500"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Add use-local-scheduler spec"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Look at simplifying the number of services deployers need to worry about"},{"line_number":10,"context_line":"when deploying Nova."}],"source_content_type":"text/x-gerrit-commit-message","patch_set":11,"id":"bff0334d_a24665bf","line":7,"range":{"start_line":7,"start_character":4,"end_line":7,"end_character":23},"updated":"2017-04-10 14:28:22.000000000","message":"rename","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":14511,"name":"Maciej Szankin","email":"maciej+review@szankin.pl","username":"mszankin"},"change_message_id":"1d94419470abb0ed2a444f7b67f7f086a4979f69","unresolved":false,"context_lines":[{"line_number":9,"context_line":"Look at simplifying the number of services deployers need to worry about"},{"line_number":10,"context_line":"when deploying Nova."},{"line_number":11,"context_line":""},{"line_number":12,"context_line":"blueprint use-conductor-scheduler"},{"line_number":13,"context_line":""},{"line_number":14,"context_line":"Change-Id: I6ced22cb066a2093711d3814a07513e595f11843"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":11,"id":"bff0334d_824361af","line":12,"range":{"start_line":12,"start_character":10,"end_line":12,"end_character":33},"updated":"2017-04-10 14:28:22.000000000","message":"rename","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"}],"specs/backlog/approved/use-conductor-scheduler.rst":[{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"76ceff0af1241ad28671239a18cb3aacb7cde157","unresolved":false,"context_lines":[{"line_number":25,"context_line":"``nova-scheduler`` and put it in ``nova-conductor``, reducing the number of"},{"line_number":26,"context_line":"processes that nova runs."},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Previously we needed to independently scale nova-conductor and nova-scheduler,"},{"line_number":29,"context_line":"but now placement API is doing most of the DB work the scheduler previously"},{"line_number":30,"context_line":"had to do. Also the benefit from running a single scheduler process is much"},{"line_number":31,"context_line":"less because of the parallelism you get inside placement API."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":11,"id":"bff0334d_029cf15a","line":30,"range":{"start_line":28,"start_character":0,"end_line":30,"end_character":10},"updated":"2017-04-10 14:23:05.000000000","message":"Until we have claims in the scheduler this is all premature. What I\u0027m specifically thinking of, is the upcall code for instance info from all computes to the host manager in the scheduler which tracks the instances on each host in a cache in the host manager and the affinity/anti-affinity filters and TypeAffinityFilter use that cache to determine placement. Until that stuff is solved with placement, I think this is premature.","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"b55834366360920547a1a59d05c140a536e275d0","unresolved":false,"context_lines":[{"line_number":25,"context_line":"``nova-scheduler`` and put it in ``nova-conductor``, reducing the number of"},{"line_number":26,"context_line":"processes that nova runs."},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Previously we needed to independently scale nova-conductor and nova-scheduler,"},{"line_number":29,"context_line":"but now placement API is doing most of the DB work the scheduler previously"},{"line_number":30,"context_line":"had to do. Also the benefit from running a single scheduler process is much"},{"line_number":31,"context_line":"less because of the parallelism you get inside placement API."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":11,"id":"bff0334d_e2dc6de5","line":30,"range":{"start_line":28,"start_character":0,"end_line":30,"end_character":10},"in_reply_to":"bff0334d_029cf15a","updated":"2017-04-10 14:32:23.000000000","message":"For more info:\n\nhttps://blueprints.launchpad.net/nova/+spec/put-host-manager-instance-info-on-a-diet","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"ce8c4222d339830121a21fd79703a51604d24c44","unresolved":false,"context_lines":[{"line_number":28,"context_line":"Previously we needed to independently scale nova-conductor and nova-scheduler,"},{"line_number":29,"context_line":"but now placement API is doing most of the DB work the scheduler previously"},{"line_number":30,"context_line":"had to do. Also the benefit from running a single scheduler process is much"},{"line_number":31,"context_line":"less because of the parallelism you get inside placement API."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Use Cases"},{"line_number":34,"context_line":"---------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"dfeb2761_6915199c","line":31,"updated":"2017-04-06 15:00:58.000000000","message":"I would talk about all the difference logs you need to consider looking through when debugging things.\n\nI think having less logs to track a single request through might actually help, although some prefer the single responsibly of the scheduler, its just some of the responsibility is now in placement.","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"ce8c4222d339830121a21fd79703a51604d24c44","unresolved":false,"context_lines":[{"line_number":38,"context_line":"* Single cell cells v2 setup, small number of compute nodes with a single"},{"line_number":39,"context_line":"  set of services."},{"line_number":40,"context_line":"* Multiple cell cells v2 setup, many compute nodes."},{"line_number":41,"context_line":"* Legacy cells v1 setup, running caching scheduler driver with a multiple"},{"line_number":42,"context_line":"  services and databases for each cell. The flow in this setup will not be"},{"line_number":43,"context_line":"  affected by proposed change."},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"Proposed change"}],"source_content_type":"text/x-rst","patch_set":11,"id":"dfeb2761_3e99891e","line":42,"range":{"start_line":41,"start_character":58,"end_line":42,"end_character":10},"updated":"2017-04-06 15:00:58.000000000","message":"normally its one nova-scheduler per cell.","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"ce8c4222d339830121a21fd79703a51604d24c44","unresolved":false,"context_lines":[{"line_number":56,"context_line":"* ``use_conductor_scheduler\u003dTrue`` (default) - create a scheduler api that"},{"line_number":57,"context_line":"  calls the manager code directly."},{"line_number":58,"context_line":"* ``use_conductor_scheduler\u003dFalse`` - create a scheduler api that calls the"},{"line_number":59,"context_line":"  manager using RPC."},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"This means the scheduler code now runs inline inside ``nova-conductor``, where"},{"line_number":62,"context_line":"all calls to the scheduler are made."}],"source_content_type":"text/x-rst","patch_set":11,"id":"bff0334d_159fed03","line":59,"updated":"2017-04-06 15:00:58.000000000","message":"Need more details on how that works, extra API in nova-conductor manager?","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"ce8c4222d339830121a21fd79703a51604d24c44","unresolved":false,"context_lines":[{"line_number":91,"context_line":"--------------------"},{"line_number":92,"context_line":""},{"line_number":93,"context_line":"Notification still appear to come from the scheduler as before, until we"},{"line_number":94,"context_line":"remove the scheduler all together."},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"Other end user impact"},{"line_number":97,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":11,"id":"dfeb2761_a992019d","line":94,"updated":"2017-04-06 15:00:58.000000000","message":"Thats probably confusing. We should use the correct host for the notifications really.","commit_id":"c210b43de72becc3af1c902d2efb21c42f895017"}],"specs/pike/approved/use-local-scheduler.rst":[{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a9ed679160649a700ae351989cdf28ae005e53d9","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Here are the following deployments to consider:"},{"line_number":34,"context_line":""},{"line_number":35,"context_line":"* Single cell cells v2 setup, small number of compute nodes"},{"line_number":36,"context_line":"* Multiple cell cells v2 setup, many compute nodes"},{"line_number":37,"context_line":"* Legacy cells v1 setup, running caching scheduler driver"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"Proposed change"},{"line_number":40,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"da36d5c6_dd6142d5","line":37,"range":{"start_line":35,"start_character":0,"end_line":37,"end_character":57},"updated":"2017-02-28 12:24:11.000000000","message":"I need to document these better.\n\nLets be explicit about the number of DBs, rabbits, nova-conductor services, etc for all these cases.","commit_id":"6a9b2000fe34ebfb19543412ed95506fb32b0026"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"4f294298e9ebc801c74ac20929e77c3f51756558","unresolved":false,"context_lines":[{"line_number":11,"context_line":"https://blueprints.launchpad.net/nova/+spec/use-local-scheduler"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"Nova has a lot of separate workers, particular when considering a full cells"},{"line_number":14,"context_line":"deployment, even with cells v2 there are multiple nova-compute process."},{"line_number":15,"context_line":""},{"line_number":16,"context_line":"This spec looks at simplifying the deployment for the majority of deployments."},{"line_number":17,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_55e258a0","line":14,"range":{"start_line":14,"start_character":55,"end_line":14,"end_character":62},"updated":"2017-03-20 15:23:02.000000000","message":"conductor.","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a301c772245d787615ac1486f4ce5f2a383c9093","unresolved":false,"context_lines":[{"line_number":19,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":20,"context_line":""},{"line_number":21,"context_line":"Make Nova easier to deploy by reducing the number of services you need to run."},{"line_number":22,"context_line":"This is more about the problems we no longer solve by separating conductor and"},{"line_number":23,"context_line":"scheduler."},{"line_number":24,"context_line":""},{"line_number":25,"context_line":"Previously we needed to independently scale nova-conductor and nova-scheduler,"},{"line_number":26,"context_line":"but now placement API is doing most of the DB work the scheduler previously"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_f7ea76ab","line":23,"range":{"start_line":22,"start_character":0,"end_line":23,"end_character":10},"updated":"2017-03-20 10:53:16.000000000","message":"this seems to confuse me a bit when reading though, not 100% sure why.","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a301c772245d787615ac1486f4ce5f2a383c9093","unresolved":false,"context_lines":[{"line_number":43,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":44,"context_line":""},{"line_number":45,"context_line":"For a legacy cells v1 setup, we keep the same number of processes they have"},{"line_number":46,"context_line":"today. Running the caching scheduler, you want a single scheduler process"},{"line_number":47,"context_line":"per cell. For the moment that also bypasses the placement API. This setup"},{"line_number":48,"context_line":"should be now deprecated, so when we drop cells v1, it can be dropped as well."},{"line_number":49,"context_line":""}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_57f34a9e","line":46,"range":{"start_line":46,"start_character":19,"end_line":46,"end_character":36},"updated":"2017-03-20 10:53:16.000000000","message":"I think send you down a strange direction here.\n\nLets catch up in person to go through the options here.\n\nIf you use placement, the separate scheduler is no longer needed.\n\nIf you use the caching scheduler, you still need a separate scheduler process.","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a301c772245d787615ac1486f4ce5f2a383c9093","unresolved":false,"context_lines":[{"line_number":53,"context_line":"* ``use_local_scheduler\u003dTrue`` (default) - create a scheduler api that calls"},{"line_number":54,"context_line":"  the manager code directly."},{"line_number":55,"context_line":"* ``use_local_scheduler\u003dFalse`` - create a scheduler api that calls the"},{"line_number":56,"context_line":"  manager using RPC. Also log a warning message that alerts the operator of"},{"line_number":57,"context_line":"  not using local scheduling."},{"line_number":58,"context_line":""},{"line_number":59,"context_line":"This means the scheduler code now runs inline inside nova-conductor, where"},{"line_number":60,"context_line":"all calls to the scheduler are made."}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_7745a6b1","line":57,"range":{"start_line":56,"start_character":21,"end_line":57,"end_character":29},"updated":"2017-03-20 10:53:16.000000000","message":"Why is there a need for a warning? What should they do? I think you mean deprecate that, I think we should do that much later, see work items. No need to log because the config will deal with that all naturally.","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a301c772245d787615ac1486f4ce5f2a383c9093","unresolved":false,"context_lines":[{"line_number":59,"context_line":"This means the scheduler code now runs inline inside nova-conductor, where"},{"line_number":60,"context_line":"all calls to the scheduler are made."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Other services"},{"line_number":63,"context_line":"--------------"},{"line_number":64,"context_line":""},{"line_number":65,"context_line":"In addition, we should also remove ``nova-cert`` and ``nova-console-auth``."},{"line_number":66,"context_line":"They are to ensure the load is removed from the API, but we can do that in the"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_d7357a46","line":63,"range":{"start_line":62,"start_character":0,"end_line":63,"end_character":14},"updated":"2017-03-20 10:53:16.000000000","message":"I would drop this one, its different to the usual template, not not really needed here.","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":782,"name":"John Garbutt","email":"john@johngarbutt.com","username":"johngarbutt"},"change_message_id":"a301c772245d787615ac1486f4ce5f2a383c9093","unresolved":false,"context_lines":[{"line_number":70,"context_line":"the ``nova-cert`` API when ``cert_use_conductor \u003d True``"},{"line_number":71,"context_line":""},{"line_number":72,"context_line":"In a similar way, when ``console_auth_use_conductor \u003d True`` will call the"},{"line_number":73,"context_line":"conductor to run the usual code, rather than calling ``console_auth``."},{"line_number":74,"context_line":""},{"line_number":75,"context_line":"Alternatives"},{"line_number":76,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":7,"id":"1a1ced50_5778aaf2","line":73,"updated":"2017-03-20 10:53:16.000000000","message":"Is this all safe to do, did we miss any here?","commit_id":"9761376150d658b2a5506500636d4833a6de6428"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"a4bff7dcb83613675baf9f03e73db998303ffb04","unresolved":false,"context_lines":[{"line_number":26,"context_line":"processes that nova runs."},{"line_number":27,"context_line":""},{"line_number":28,"context_line":"Previously we needed to independently scale nova-conductor and nova-scheduler,"},{"line_number":29,"context_line":"but now placement API is doing most of the DB work the scheduler previously"},{"line_number":30,"context_line":"had to do. Also the benefit from running a single scheduler process is much"},{"line_number":31,"context_line":"less because of the parallelism you get inside placement API."},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffe62b97_8d6f3b91","line":30,"range":{"start_line":29,"start_character":4,"end_line":30,"end_character":10},"updated":"2017-03-27 15:56:52.000000000","message":"Fairly true, but I\u0027d love to see the spec commenting what\u0027s still used by the scheduler, and how we could use the conductor for that, ie.\n- filters/weighters running\n- scheduler consuming in-memory \"HostStates\" - which is a concurrency problem for the scheduler process being A/A HA.\n- computes doing RPC fanouts to the scheduler for providing resources for aggregates and instances. That\u0027s something the placement API service could help, but that\u0027s not something yet there.","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"a4bff7dcb83613675baf9f03e73db998303ffb04","unresolved":false,"context_lines":[{"line_number":37,"context_line":""},{"line_number":38,"context_line":"* Single cell cells v2 setup, small number of compute nodes with a single"},{"line_number":39,"context_line":"  set of services."},{"line_number":40,"context_line":"* Multiple cell cells v2 setup, many compute nodes."},{"line_number":41,"context_line":"* Legacy cells v1 setup, running caching scheduler driver with a multiple"},{"line_number":42,"context_line":"  services and databases for each cell. The flow in this setup will not be"},{"line_number":43,"context_line":"  affected by proposed change."}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffe62b97_adaf9fe1","line":40,"updated":"2017-03-27 15:56:52.000000000","message":"So, we agreed previously that for cellsv2, a global scheduler is something we wanted.\nGiven we\u0027d like to merge it with the conductor, I want to make it cleat that then it would be merged with the new \"super-conductor\" service which would be global.","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"625794e374231b981f82f481187fdc701e59155f","unresolved":false,"context_lines":[{"line_number":47,"context_line":""},{"line_number":48,"context_line":"For a legacy cells v1 setup, we keep the same number of processes they have"},{"line_number":49,"context_line":"today. Running the caching scheduler, you want a single scheduler process"},{"line_number":50,"context_line":"per cell. For the moment that also bypasses the placement API. This setup"},{"line_number":51,"context_line":"should be now deprecated, so when we drop cells v1, it can be dropped as well."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"To alternate between different setups, the ``use_local_scheduler`` option"},{"line_number":54,"context_line":"can be introduced:"}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffe62b97_22515f0f","line":51,"range":{"start_line":50,"start_character":63,"end_line":51,"end_character":24},"updated":"2017-03-27 14:31:59.000000000","message":"As far as I know, the caching scheduler is not yet deprecated, or this hasn\u0027t been actually signaled in the release notes or docs.","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"},{"author":{"_account_id":11564,"name":"Chris Dent","email":"cdent@anticdent.org","username":"chdent"},"change_message_id":"b6fccda541924bc41c2dfbd99b5f515247aaeb6c","unresolved":false,"context_lines":[{"line_number":51,"context_line":"should be now deprecated, so when we drop cells v1, it can be dropped as well."},{"line_number":52,"context_line":""},{"line_number":53,"context_line":"To alternate between different setups, the ``use_local_scheduler`` option"},{"line_number":54,"context_line":"can be introduced:"},{"line_number":55,"context_line":""},{"line_number":56,"context_line":"* ``use_local_scheduler\u003dTrue`` (default) - create a scheduler api that calls"},{"line_number":57,"context_line":"  the manager code directly."}],"source_content_type":"text/x-rst","patch_set":9,"id":"dfeb2761_2079a8fb","line":54,"updated":"2017-03-31 13:25:49.000000000","message":"\u0027local\u0027 seems like a bad name for this, as it doesn\u0027t really mean local, it means \"in the conductor\"","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"a4bff7dcb83613675baf9f03e73db998303ffb04","unresolved":false,"context_lines":[{"line_number":59,"context_line":"  manager using RPC."},{"line_number":60,"context_line":""},{"line_number":61,"context_line":"This means the scheduler code now runs inline inside ``nova-conductor``, where"},{"line_number":62,"context_line":"all calls to the scheduler are made."},{"line_number":63,"context_line":""},{"line_number":64,"context_line":"Alternatives"},{"line_number":65,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffe62b97_789397a4","line":62,"updated":"2017-03-27 15:56:52.000000000","message":"So, say you agree with me on having the scheduler modules now provided by the super-conductor RPC service. Given the computes are atm RPC-casting the scheduler MQ, which would then be an upcall, how do you think we could use the placement service for that ?","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"625794e374231b981f82f481187fdc701e59155f","unresolved":false,"context_lines":[{"line_number":101,"context_line":"Performance Impact"},{"line_number":102,"context_line":"------------------"},{"line_number":103,"context_line":""},{"line_number":104,"context_line":"You may now need additional API cell nova-conductor processes to take the"},{"line_number":105,"context_line":"load that was previously on the scheduler process."},{"line_number":106,"context_line":""},{"line_number":107,"context_line":"Other deployer impact"},{"line_number":108,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":9,"id":"ffe62b97_e2b937e9","line":105,"range":{"start_line":104,"start_character":0,"end_line":105,"end_character":50},"updated":"2017-03-27 14:31:59.000000000","message":"I\u0027ve heard large ops complaints about how big nova-conductor needs to be and resources it consumes, given all of the different hats it wears and work going through it, adding more things to nova-conductor seems like maybe a bad idea given the existing load on n-cond, eventually making it a single point of failure (although you\u0027d scale out multiples of them horizontally, so it\u0027s less of a single point of failure, and more like a handful). Anyway, it\u0027s something worth considering. In Laski\u0027s brain dump on cells v2 before he left he actually talked about how overly-complicated conductor was getting and all of the roles it takes on, and we might consider splitting it up, which would be opposite to this idea.\n\nhttps://review.openstack.org/#/c/390660/1/doc/source/cells.rst@395","commit_id":"50e9557cb37d29a4941e00bc8c59446b44b5a1e6"}]}