)]}'
{"/COMMIT_MSG":[{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"ffc0cf591b764b213d4cbcb869a028c4db1d99f5","unresolved":false,"context_lines":[{"line_number":4,"context_line":"Commit:     Nikola Dipanov \u003cndipanov@redhat.com\u003e"},{"line_number":5,"context_line":"CommitDate: 2015-06-15 15:29:17 +0100"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Scheduler Introduce lightwieght transactional model for HostState"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Blueprint: host-state-level-locking"},{"line_number":10,"context_line":"Change-Id: Iedb4291801e4dc159ad8aacbaf3a602512c87868"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"fa32b979_8de4d4ac","line":7,"updated":"2015-06-19 17:35:36.000000000","message":"s/lightwieght/lightweight/","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":2271,"name":"Michael Still","email":"mikal@stillhq.com","username":"mikalstill"},"change_message_id":"87270cf04b3217670bb49255c249f57f8556c23c","unresolved":false,"context_lines":[{"line_number":4,"context_line":"Commit:     Nikola Dipanov \u003cndipanov@redhat.com\u003e"},{"line_number":5,"context_line":"CommitDate: 2015-06-15 15:29:17 +0100"},{"line_number":6,"context_line":""},{"line_number":7,"context_line":"Scheduler Introduce lightwieght transactional model for HostState"},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Blueprint: host-state-level-locking"},{"line_number":10,"context_line":"Change-Id: Iedb4291801e4dc159ad8aacbaf3a602512c87868"}],"source_content_type":"text/x-gerrit-commit-message","patch_set":5,"id":"fa32b979_5d1ad718","line":7,"in_reply_to":"fa32b979_8de4d4ac","updated":"2015-06-22 10:23:18.000000000","message":"Yeah, that thing.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"}],"specs/liberty/approved/host-state-level-locking.rst":[{"author":{"_account_id":7730,"name":"Sahid Orentino Ferdjaoui","email":"sahid.ferdjaoui@industrialdiscipline.com","username":"sahid"},"change_message_id":"6cfe0b098d81301d19a2ebb42e0ffecccb345550","unresolved":false,"context_lines":[{"line_number":82,"context_line":""},{"line_number":83,"context_line":"Firstly, it would be very useful to use the Claim logic instead (or inside) of"},{"line_number":84,"context_line":"HostState.consume_from_instance() as there is almost complete duplication"},{"line_number":85,"context_line":"there."},{"line_number":86,"context_line":""},{"line_number":87,"context_line":"Next change that would be in the scope for this blueprint would be adding"},{"line_number":88,"context_line":"synchronisation primitives around accessing and updating HostState fields."}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_4df3210b","line":85,"updated":"2015-06-15 08:39:45.000000000","message":"I guess it is clearly the way we have to follow, I wondering to know if you can give more information about the way you will achieve this (about the implementation).","commit_id":"1cabaf0c64b102067845f32ecc2e1da579bd3949"},{"author":{"_account_id":7730,"name":"Sahid Orentino Ferdjaoui","email":"sahid.ferdjaoui@industrialdiscipline.com","username":"sahid"},"change_message_id":"6cfe0b098d81301d19a2ebb42e0ffecccb345550","unresolved":false,"context_lines":[{"line_number":92,"context_line":"is synchronized, as once the host is chosen, it will need to have resources"},{"line_number":93,"context_line":"consumed (remember - many concurrent threads could be trying to consume"},{"line_number":94,"context_line":"resources from the same HostState) and if it fails any of the claims, no"},{"line_number":95,"context_line":"resources should be consumed."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":98,"context_line":"to take into account the failure to claim in consume_from_instance() and try"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_adbf4dcf","line":95,"updated":"2015-06-15 08:39:45.000000000","message":"Just for me to understand how the synchronization will work. If i understand well we can have for host A a HostState in scheduler1 and another in scheduler2, right? Or are you talking about a local HostState which has to support contention between multiple requests (threads)?","commit_id":"1cabaf0c64b102067845f32ecc2e1da579bd3949"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"2f0bf31ef38f5f3cc6d465c27b6132578700cd3e","unresolved":false,"context_lines":[{"line_number":92,"context_line":"is synchronized, as once the host is chosen, it will need to have resources"},{"line_number":93,"context_line":"consumed (remember - many concurrent threads could be trying to consume"},{"line_number":94,"context_line":"resources from the same HostState) and if it fails any of the claims, no"},{"line_number":95,"context_line":"resources should be consumed."},{"line_number":96,"context_line":""},{"line_number":97,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":98,"context_line":"to take into account the failure to claim in consume_from_instance() and try"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_5b8f5823","line":95,"in_reply_to":"fa32b979_adbf4dcf","updated":"2015-06-15 10:41:58.000000000","message":"Yeah so I will add a note about this. This does not take into account the case of several scheduler processes. \n\nThe way to fix that would be to have a DistributedHostState, where the locking is distributed (i.e. instead of in-memory, the state of the hosts is in a Zookeeper cluster).\n\nThis is outside of the scope of this BP, but I will make a note for clarity.","commit_id":"1cabaf0c64b102067845f32ecc2e1da579bd3949"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"ea45b2e60bac80b28c789ac90ea796224063e8a0","unresolved":false,"context_lines":[{"line_number":173,"context_line":"* Modify HostState.consume_from_instance() to use the Claim logic"},{"line_number":174,"context_line":""},{"line_number":175,"context_line":"* Modify FilterSchedule._schedule() method to expect a claim transaction"},{"line_number":176,"context_line":"  failure and take appropriate action."},{"line_number":177,"context_line":""},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"Dependencies"}],"source_content_type":"text/x-rst","patch_set":3,"id":"1af86dd1_99a5fd05","line":176,"updated":"2015-06-12 21:53:05.000000000","message":"fun : https://etherpad.openstack.org/p/liberty-nova-scalable-scheduler\n\nUsing Claims is good IMHO, but I\u0027ll review your spec next week.","commit_id":"1cabaf0c64b102067845f32ecc2e1da579bd3949"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"f456f0658301ecd5ba6efcf4f95eb44a423082d2","unresolved":false,"context_lines":[{"line_number":173,"context_line":"* Modify HostState.consume_from_instance() to use the Claim logic"},{"line_number":174,"context_line":""},{"line_number":175,"context_line":"* Modify FilterSchedule._schedule() method to expect a claim transaction"},{"line_number":176,"context_line":"  failure and take appropriate action."},{"line_number":177,"context_line":""},{"line_number":178,"context_line":""},{"line_number":179,"context_line":"Dependencies"}],"source_content_type":"text/x-rst","patch_set":3,"id":"fa32b979_07dc18e3","line":176,"in_reply_to":"1af86dd1_99a5fd05","updated":"2015-06-15 07:48:39.000000000","message":"Right so I reviewed that and this proposal is not it - it just proposes very light weight locking with retries, while John\u0027s design would get rid of the 2PC kind of approach we have now as we would be doing the ComputeNode updates in the scheduler too.\n\nUsing claims here literally means - using the same code but instead of updating the compute_node table, we update the HostState object in scheduler memory. I should probably clarify.","commit_id":"1cabaf0c64b102067845f32ecc2e1da579bd3949"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"61cc6588063119354b8af438d6b30e4a7c56c85e","unresolved":false,"context_lines":[{"line_number":31,"context_line":""},{"line_number":32,"context_line":"Our scheduler service is inherently multi-threaded as it currently runs an"},{"line_number":33,"context_line":"oslo-messaging RpcServer using an EventletExecutor. This means that every"},{"line_number":34,"context_line":"incoming RPC message for select_destinations will be dispatched in it\u0027s own"},{"line_number":35,"context_line":"green thread."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Upon receiving the message, every green thread will read all ComputeNode states"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_c6ff0000","line":34,"updated":"2015-06-15 14:46:16.000000000","message":"micro-nit: s/it\u0027s/its","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"467e95dc7009ba9f57299d4a859e07b95c2c2394","unresolved":false,"context_lines":[{"line_number":61,"context_line":"   It is worth noting here that even though we may find that host seems like"},{"line_number":62,"context_line":"   it will be failing, we may still want to choose it, as we don\u0027t ever drop"},{"line_number":63,"context_line":"   the resources consumed on the HostState even after we register a retry from"},{"line_number":64,"context_line":"   a already chosen compute host in this refresh cycle, so it may in fact be"},{"line_number":65,"context_line":"   a false negative."},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"2. There needs to be some kind of locking that is granular enough so as not to"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_cda2dc9a","line":64,"updated":"2015-06-19 17:34:51.000000000","message":"nit: s/a already chosen/an already-chosen/","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":8276,"name":"kiwik","email":"chenrui.momo@gmail.com","username":"kiwik"},"change_message_id":"30c64de80236f59b68e848d33400d565b857a3c6","unresolved":false,"context_lines":[{"line_number":85,"context_line":"Proposed change"},{"line_number":86,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":87,"context_line":""},{"line_number":88,"context_line":"Firstly, it would be very useful to use the Claim logic instead (or inside) of"},{"line_number":89,"context_line":"HostState.consume_from_instance() as there is almost complete duplication"},{"line_number":90,"context_line":"there."},{"line_number":91,"context_line":""}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_4da3ba00","line":88,"updated":"2015-06-17 04:15:58.000000000","message":"I like this idea.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"61cc6588063119354b8af438d6b30e4a7c56c85e","unresolved":false,"context_lines":[{"line_number":90,"context_line":"there."},{"line_number":91,"context_line":""},{"line_number":92,"context_line":"Next change that would be in the scope for this blueprint would be adding"},{"line_number":93,"context_line":"synchronisation primitives around accessing and updating HostState fields."},{"line_number":94,"context_line":"A lightweight approach would be to not use any synchronisation primitives in"},{"line_number":95,"context_line":"the filters, as access to the host state is a) read-only b) usually per"},{"line_number":96,"context_line":"resource. consume_from_instance is the place where we want to make sure access"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_0635e8d8","line":93,"updated":"2015-06-15 14:46:16.000000000","message":"I\u0027m not seeing how this would help. Each thread has its own collection of HostState objects in memory, no? How will locking one thread\u0027s update of that object\u0027s values benefit the others? They will still base their filtering decision on the values in the DB when _get_all_host_states() is called.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"6f5c92cb3760778f5d1ed1358dbb210812d17df8","unresolved":false,"context_lines":[{"line_number":90,"context_line":"there."},{"line_number":91,"context_line":""},{"line_number":92,"context_line":"Next change that would be in the scope for this blueprint would be adding"},{"line_number":93,"context_line":"synchronisation primitives around accessing and updating HostState fields."},{"line_number":94,"context_line":"A lightweight approach would be to not use any synchronisation primitives in"},{"line_number":95,"context_line":"the filters, as access to the host state is a) read-only b) usually per"},{"line_number":96,"context_line":"resource. consume_from_instance is the place where we want to make sure access"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_47f68790","line":93,"in_reply_to":"fa32b979_0635e8d8","updated":"2015-06-15 16:55:30.000000000","message":"Nope - it\u0027s part of the HostManager.host_state_map which is instantianed once on startup and share among all the threads.\n\nIt is updated from the DB and not even every time:\n\nhttps://github.com/openstack/nova/blob/master/nova/scheduler/host_manager.py#L190\n\nSo if there is a lot of requests  (high load) we are actually delaying updating.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"2d89940f5219b4c7612214310f7cd6115e57f9c2","unresolved":false,"context_lines":[{"line_number":90,"context_line":"there."},{"line_number":91,"context_line":""},{"line_number":92,"context_line":"Next change that would be in the scope for this blueprint would be adding"},{"line_number":93,"context_line":"synchronisation primitives around accessing and updating HostState fields."},{"line_number":94,"context_line":"A lightweight approach would be to not use any synchronisation primitives in"},{"line_number":95,"context_line":"the filters, as access to the host state is a) read-only b) usually per"},{"line_number":96,"context_line":"resource. consume_from_instance is the place where we want to make sure access"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_58abc889","line":93,"in_reply_to":"fa32b979_47f68790","updated":"2015-06-15 17:38:40.000000000","message":"Ah, I think I was misreading due to the emphasis on locking. I thought you were proposing locks on the DB call to ComputeNodeList.get_all(), so that the info there wouldn\u0027t be out-of-sync with the result of a claim.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"61cc6588063119354b8af438d6b30e4a7c56c85e","unresolved":false,"context_lines":[{"line_number":98,"context_line":"consumed (remember - many concurrent threads could be trying to consume"},{"line_number":99,"context_line":"resources from the same HostState) and if it fails any of the \"claims\", no"},{"line_number":100,"context_line":"resources should be consumed. Updating the host state with fresh values after"},{"line_number":101,"context_line":"a DB read should also be synchronized."},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_e679e494","line":101,"updated":"2015-06-15 14:46:16.000000000","message":"Which DB reads are you thinking of? The only one that should affect scheduler selection is the initial _get_all_host_states().","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"6f5c92cb3760778f5d1ed1358dbb210812d17df8","unresolved":false,"context_lines":[{"line_number":98,"context_line":"consumed (remember - many concurrent threads could be trying to consume"},{"line_number":99,"context_line":"resources from the same HostState) and if it fails any of the \"claims\", no"},{"line_number":100,"context_line":"resources should be consumed. Updating the host state with fresh values after"},{"line_number":101,"context_line":"a DB read should also be synchronized."},{"line_number":102,"context_line":""},{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_478767b5","line":101,"in_reply_to":"fa32b979_e679e494","updated":"2015-06-15 16:55:30.000000000","message":"right so I mean _get_all_host_states() runs and the code from the above comment decides it needs to update it in memory","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":8276,"name":"kiwik","email":"chenrui.momo@gmail.com","username":"kiwik"},"change_message_id":"30c64de80236f59b68e848d33400d565b857a3c6","unresolved":false,"context_lines":[{"line_number":102,"context_line":""},{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"},{"line_number":105,"context_line":"the next host that passed the filters, or choose to ignore the local in memory"},{"line_number":106,"context_line":"failure and risk a retry from the compute host."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"It is worth noting that this proposal only looks at fixing data consistency"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_ad0cc6ed","line":105,"updated":"2015-06-17 04:15:58.000000000","message":"How we get the next host? repick from weighed_hosts?","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"60c9fe1c6df4e084ec70f564abfa87e417e49186","unresolved":false,"context_lines":[{"line_number":102,"context_line":""},{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"},{"line_number":105,"context_line":"the next host that passed the filters, or choose to ignore the local in memory"},{"line_number":106,"context_line":"failure and risk a retry from the compute host."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"It is worth noting that this proposal only looks at fixing data consistency"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_e9ab9b73","line":105,"in_reply_to":"fa32b979_ad0cc6ed","updated":"2015-06-17 09:19:53.000000000","message":"Yes that was my idea pretty much","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":1063,"name":"Ed Leafe","email":"ed@leafe.com","username":"ed-leafe"},"change_message_id":"61cc6588063119354b8af438d6b30e4a7c56c85e","unresolved":false,"context_lines":[{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"},{"line_number":105,"context_line":"the next host that passed the filters, or choose to ignore the local in memory"},{"line_number":106,"context_line":"failure and risk a retry from the compute host."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"It is worth noting that this proposal only looks at fixing data consistency"},{"line_number":109,"context_line":"among threads of a single nova-scheduler process. Running several workers still"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_a673dcb0","line":106,"updated":"2015-06-15 14:46:16.000000000","message":"The failure doesn\u0027t come in consume_from_instance(), since that thread\u0027s copy of the host hasn\u0027t changed. It will come when the compute cannot boot the instance on the selected host, and select_destinations() is called again.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"6f5c92cb3760778f5d1ed1358dbb210812d17df8","unresolved":false,"context_lines":[{"line_number":103,"context_line":"Final piece of the puzzle is modifying the FilterScheduler._schedule() method"},{"line_number":104,"context_line":"to take into account the failure to claim in consume_from_instance() and try"},{"line_number":105,"context_line":"the next host that passed the filters, or choose to ignore the local in memory"},{"line_number":106,"context_line":"failure and risk a retry from the compute host."},{"line_number":107,"context_line":""},{"line_number":108,"context_line":"It is worth noting that this proposal only looks at fixing data consistency"},{"line_number":109,"context_line":"among threads of a single nova-scheduler process. Running several workers still"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_c7c137e6","line":106,"in_reply_to":"fa32b979_a673dcb0","updated":"2015-06-15 16:55:30.000000000","message":"We established - there is no thread\u0027s copy.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"7e837ee25047c81c6871784c8789eec5dd42727d","unresolved":false,"context_lines":[{"line_number":108,"context_line":"It is worth noting that this proposal only looks at fixing data consistency"},{"line_number":109,"context_line":"among threads of a single nova-scheduler process. Running several workers still"},{"line_number":110,"context_line":"means that their internal state is going to be inconsistent between updates"},{"line_number":111,"context_line":"from the database. Fixing this is outside of the scope of this proposal."},{"line_number":112,"context_line":""},{"line_number":113,"context_line":"Alternatives"},{"line_number":114,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_44f22c71","line":111,"updated":"2015-06-19 15:44:19.000000000","message":"Thanks for adding that paragraph, it makes the scope of that blueprint very clear. The other benefit I can see from that BP is that it puts the base blocks for having a lock which can be distributed, by reusing the Claim model.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"467e95dc7009ba9f57299d4a859e07b95c2c2394","unresolved":false,"context_lines":[{"line_number":150,"context_line":"this change which may decrease the average response time for basic workloads,"},{"line_number":151,"context_line":"I fully expect this to massively improve the performance in conditions of a"},{"line_number":152,"context_line":"large number of requests, or low overall cloud capacity (or specific resources"},{"line_number":153,"context_line":"such as Ironic hosts), as it will significantly cut down on issued retries."},{"line_number":154,"context_line":""},{"line_number":155,"context_line":"Other deployer impact"},{"line_number":156,"context_line":"---------------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_cd6c1c39","line":153,"updated":"2015-06-19 17:34:51.000000000","message":"Agreed.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":7,"name":"Jay Pipes","email":"jaypipes@gmail.com","username":"jaypipes"},"change_message_id":"467e95dc7009ba9f57299d4a859e07b95c2c2394","unresolved":false,"context_lines":[{"line_number":156,"context_line":"---------------------"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"There may be several config options deployers would need to consider. Defaults"},{"line_number":159,"context_line":"may be chosen in such a way as to not change previous behaviour."},{"line_number":160,"context_line":""},{"line_number":161,"context_line":"Developer impact"},{"line_number":162,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_6d64f048","line":159,"updated":"2015-06-19 17:34:51.000000000","message":"What configuration options are you thinking would be needed? I personally can\u0027t think of any...","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":5511,"name":"Nikola Dipanov","email":"ndipanov@redhat.com","username":"ndipanov"},"change_message_id":"d0086210cb038cd6076e169fd6df1bc3c4d1d4b0","unresolved":false,"context_lines":[{"line_number":156,"context_line":"---------------------"},{"line_number":157,"context_line":""},{"line_number":158,"context_line":"There may be several config options deployers would need to consider. Defaults"},{"line_number":159,"context_line":"may be chosen in such a way as to not change previous behaviour."},{"line_number":160,"context_line":""},{"line_number":161,"context_line":"Developer impact"},{"line_number":162,"context_line":"----------------"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_b0548f21","line":159,"in_reply_to":"fa32b979_6d64f048","updated":"2015-06-19 18:05:08.000000000","message":"Yeah that\u0027s why I kind of left it open-ended... There are 2 things that could be tweaked - 1) turning the locking off (probably useless), and turning the \"local retry\" off, in the sense that we still want to go ahead and choose the host even though we think that it will fail. Also probably useless come to think of it.\n\nIf none of that happens during implementation - I can just back this bit out of the spec.","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"7e837ee25047c81c6871784c8789eec5dd42727d","unresolved":false,"context_lines":[{"line_number":207,"context_line":""},{"line_number":208,"context_line":"There may be an additional config option to turn on the transactional nature"},{"line_number":209,"context_line":"of consume_from_instance() and possibly another one to tell the scheduler to"},{"line_number":210,"context_line":"go ahead and attempt to land an instance even though a local claim failed."},{"line_number":211,"context_line":""},{"line_number":212,"context_line":"References"},{"line_number":213,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":5,"id":"fa32b979_a40bf82a","line":210,"updated":"2015-06-19 15:44:19.000000000","message":"I guess the default config option will be to enable the transactions, right?","commit_id":"2c7831fb5a3c25c07993fc3559b1d9989c1a51c5"}]}
