)]}'
{"specs/liberty/approved/robustify_evacuate.rst":[{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"26f90519fef5727ccb08a75975a959cb09978737","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Include the URL of your launchpad blueprint:"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"https://blueprints.launchpad.net/nova/+spec/example"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Evacuate provides the ability for a deployer to relocate and rebuild"},{"line_number":16,"context_line":"instances when a compute host goes down for an extended period of"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_e62a402c","line":13,"updated":"2015-04-29 20:01:49.000000000","message":"https://blueprints.launchpad.net/nova/+spec/robustify_evacuate","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"5dcca6e2c9e21ec90cf9383667d78526107f99f6","unresolved":false,"context_lines":[{"line_number":10,"context_line":""},{"line_number":11,"context_line":"Include the URL of your launchpad blueprint:"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"https://blueprints.launchpad.net/nova/+spec/example"},{"line_number":14,"context_line":""},{"line_number":15,"context_line":"Evacuate provides the ability for a deployer to relocate and rebuild"},{"line_number":16,"context_line":"instances when a compute host goes down for an extended period of"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_a9c1bb82","line":13,"in_reply_to":"3acd31a7_e62a402c","updated":"2015-04-29 20:12:17.000000000","message":"No, really, it\u0027s called example!\n\nEr, oops :)","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":2271,"name":"Michael Still","email":"mikal@stillhq.com","username":"mikalstill"},"change_message_id":"4c46c3327c9f6904f889987c02180195957fe5f1","unresolved":false,"context_lines":[{"line_number":31,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_c0080b30","line":34,"range":{"start_line":34,"start_character":34,"end_line":34,"end_character":45},"updated":"2015-04-24 03:12:02.000000000","message":"relatively","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"26f90519fef5727ccb08a75975a959cb09978737","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"},{"line_number":38,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_869f0c5f","line":35,"updated":"2015-04-29 20:01:49.000000000","message":"The guess is usually based on a decent assumption, but it might be good to point out that there\u0027s real danger in using evacuate with a clustered hypervisor.  Or a renegade HA compute setup.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"bcda605f81f35d868cbd9e168c91fcee8344eb78","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"},{"line_number":38,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_b57c6821","line":35,"in_reply_to":"3acd31a7_0cdd5d42","updated":"2015-04-29 22:24:20.000000000","message":"Alright, I concede that it\u0027s not a decent assumption.  A changing hostname will break things, but it losing data is very bad.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"2f149b003141a67cacdc272fff6b6208d4bdb45b","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"},{"line_number":38,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_0cdd5d42","line":35,"in_reply_to":"3acd31a7_6cb6113f","updated":"2015-04-29 20:41:30.000000000","message":"I don\u0027t mean changed intentionally. I mean, you reboot your compute host, it fails to look up its hostname in DNS, or never gets a DHCP reply, defaults to calling itself \"localhost\" and then when nova starts, your instances get deleted.\n\nAnother case: let\u0027s say you have a hardware failure. You pull the disks from the dead machine, put them into a good identical machine and boot. The MAC on the NIC is different, dhcp gives you a different hostname, you continue booting and *BAM* instances gone.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"5dcca6e2c9e21ec90cf9383667d78526107f99f6","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"},{"line_number":38,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_e9ce93b3","line":35,"in_reply_to":"3acd31a7_869f0c5f","updated":"2015-04-29 20:12:17.000000000","message":"It\u0027s really not a decent assumption I don\u0027t think. There\u0027s nothing in the evaluation that includes \"has evacuate ever been called on this instance\". With libvirt, you can simply change your system hostname and nova will delete all your instances.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"5681bb08804f57cf3e53bfe295bb433e98b511a2","unresolved":false,"context_lines":[{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"},{"line_number":38,"context_line":"----------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_6cb6113f","line":35,"in_reply_to":"3acd31a7_e9ce93b3","updated":"2015-04-29 20:36:05.000000000","message":"Fair.  But if you changed the hostname would any calls make it to the compute for that instance, isn\u0027t RPC keyed off of host?  My point just being that I would expect changing the hostname would break a bunch of other things as well, and shouldn\u0027t be done.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"26f90519fef5727ccb08a75975a959cb09978737","unresolved":false,"context_lines":[{"line_number":38,"context_line":"----------"},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"A developer wants to enable evacuation on their cloud and have Nova"},{"line_number":41,"context_line":"only destroy data when it has been safelt moved to another host."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Project Priority"},{"line_number":44,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_465e94ea","line":41,"range":{"start_line":41,"start_character":35,"end_line":41,"end_character":41},"updated":"2015-04-29 20:01:49.000000000","message":"safely","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":2271,"name":"Michael Still","email":"mikal@stillhq.com","username":"mikalstill"},"change_message_id":"4c46c3327c9f6904f889987c02180195957fe5f1","unresolved":false,"context_lines":[{"line_number":66,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":67,"context_line":"current migration object so that we can keep user/admin-initiated"},{"line_number":68,"context_line":"migration records separate from evacuation-initiated ones (when and"},{"line_number":69,"context_line":"where appropriate)."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"Alternatives"},{"line_number":72,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_fb6980bf","line":69,"updated":"2015-04-24 03:12:02.000000000","message":"Is this sufficient? What if the compute node which is evacuated from is down for a long time and comes back after a second evacuation? Something like:\n\n - compute A is evacuated, instance moves to compute B\n - compute B is evacuated, instance moves to compute C\n - compute A comes back, migration record shows a B-\u003eC move\n\nHow do we handle that situation?","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"b6cf0d99b83f381931ee6628b80f58bfd210073b","unresolved":false,"context_lines":[{"line_number":66,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":67,"context_line":"current migration object so that we can keep user/admin-initiated"},{"line_number":68,"context_line":"migration records separate from evacuation-initiated ones (when and"},{"line_number":69,"context_line":"where appropriate)."},{"line_number":70,"context_line":""},{"line_number":71,"context_line":"Alternatives"},{"line_number":72,"context_line":"------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_d4c0a62c","line":69,"in_reply_to":"bab921fb_fb6980bf","updated":"2015-04-24 15:22:38.000000000","message":"There would be two records in that case, one for the A-B move and one for B-C, right? The first is a tombstone that only A can clean up, the second is for B. Right now the way we handle migration records (for actual migrations) is that they remain in the database in complete state after it\u0027s done, right? So this becomes a log, and we can use that log to signal to any of the source hosts along the way in specific order I think.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"26f90519fef5727ccb08a75975a959cb09978737","unresolved":false,"context_lines":[{"line_number":91,"context_line":"This would re-use our existing migration data structure, so data model"},{"line_number":92,"context_line":"impact would be minor. We would need to add at least one field to"},{"line_number":93,"context_line":"track the reason for the migration, being \"user initiated\" or \"a"},{"line_number":94,"context_line":"result of an evacuation.\""},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"REST API impact"},{"line_number":97,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_c6dbc4ec","line":94,"updated":"2015-04-29 20:01:49.000000000","message":"It might be better to record the method of migration rather than the reason.  A case could be made for an enum here, which might be a bit much.  But a method field that could be migrate/evacuate/live-migrate(eventually) would be good.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"5dcca6e2c9e21ec90cf9383667d78526107f99f6","unresolved":false,"context_lines":[{"line_number":91,"context_line":"This would re-use our existing migration data structure, so data model"},{"line_number":92,"context_line":"impact would be minor. We would need to add at least one field to"},{"line_number":93,"context_line":"track the reason for the migration, being \"user initiated\" or \"a"},{"line_number":94,"context_line":"result of an evacuation.\""},{"line_number":95,"context_line":""},{"line_number":96,"context_line":"REST API impact"},{"line_number":97,"context_line":"---------------"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_090f8fe2","line":94,"in_reply_to":"3acd31a7_c6dbc4ec","updated":"2015-04-29 20:12:17.000000000","message":"Sure, that supports John\u0027s desire to include live migrations here.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"26f90519fef5727ccb08a75975a959cb09978737","unresolved":false,"context_lines":[{"line_number":170,"context_line":"Testing this in tempest is tricky because it effectively requires"},{"line_number":171,"context_line":"downing a running compute node. Thus, functional tests will be used to"},{"line_number":172,"context_line":"spin up a reasonably full environment with multiple computes in order"},{"line_number":173,"context_line":"to test the feature."},{"line_number":174,"context_line":""},{"line_number":175,"context_line":"Documentation Impact"},{"line_number":176,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"3acd31a7_66a75050","line":173,"updated":"2015-04-29 20:01:49.000000000","message":"And testing a volume backed instance may require downing the compute host, or severing an iscsi connection via some other means.","commit_id":"4833189eb17a0ae5e0466d800eaf95c5578423b2"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"6f4bf0f31b09527ffb2ae48d4980889cec813e42","unresolved":false,"context_lines":[{"line_number":8,"context_line":"Robustify Evacuate"},{"line_number":9,"context_line":"\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":"Include the URL of your launchpad blueprint:"},{"line_number":12,"context_line":""},{"line_number":13,"context_line":"https://blueprints.launchpad.net/nova/+spec/robustify-evacuate"},{"line_number":14,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_d6e4633c","line":11,"updated":"2015-04-30 21:50:37.000000000","message":"eh, I love to nitpick that kind of miss :-)","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"6f4bf0f31b09527ffb2ae48d4980889cec813e42","unresolved":false,"context_lines":[{"line_number":76,"context_line":" 1. Check for unconfirmed evacuation migrations where it is the source"},{"line_number":77,"context_line":"    host"},{"line_number":78,"context_line":" 2. Delete the local data, if still present"},{"line_number":79,"context_line":" 3. Mark the migration as completed"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":82,"context_line":"current migration object so that we can keep user/admin-initiated"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_fccaa1a0","line":79,"updated":"2015-04-30 21:50:37.000000000","message":"++","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":5441,"name":"Andrew Laski","email":"andrew@lascii.com","username":"alaski"},"change_message_id":"9f24a577ed5ca54047f34cf80927eecff93f09e7","unresolved":false,"context_lines":[{"line_number":78,"context_line":" 2. Delete the local data, if still present"},{"line_number":79,"context_line":" 3. Mark the migration as completed"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":82,"context_line":"current migration object so that we can keep user/admin-initiated"},{"line_number":83,"context_line":"migration records separate from evacuation-initiated ones (when and"},{"line_number":84,"context_line":"where appropriate)."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_6a724ae7","line":81,"updated":"2015-04-30 19:10:46.000000000","message":"My preference is on type.  But this can be debated in code.","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"6f4bf0f31b09527ffb2ae48d4980889cec813e42","unresolved":false,"context_lines":[{"line_number":78,"context_line":" 2. Delete the local data, if still present"},{"line_number":79,"context_line":" 3. Mark the migration as completed"},{"line_number":80,"context_line":""},{"line_number":81,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":82,"context_line":"current migration object so that we can keep user/admin-initiated"},{"line_number":83,"context_line":"migration records separate from evacuation-initiated ones (when and"},{"line_number":84,"context_line":"where appropriate)."}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_5ce3d533","line":81,"in_reply_to":"3acd31a7_6a724ae7","updated":"2015-04-30 21:50:37.000000000","message":"+1","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"6f4bf0f31b09527ffb2ae48d4980889cec813e42","unresolved":false,"context_lines":[{"line_number":105,"context_line":"those fields. However, both of these approaches suffer from the fact"},{"line_number":106,"context_line":"that they\u0027re guessing based on circumstantial data instead of a record"},{"line_number":107,"context_line":"that tells us an evacuation was performed."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"Data model impact"},{"line_number":110,"context_line":"-----------------"},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_3c2db946","line":108,"updated":"2015-04-30 21:50:37.000000000","message":"Maybe another alternative would be to mark all evacuated instances with a flag like \u0027evacuated_from\u003dhost|None\u0027 and figure out at the startup which local instances are null for that field, but that just sounds another bad alternative since I prefer your approach so it\u0027s maybe not needed to put that in the section\n\nDo I overthink ? Yes.","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"b0b36b9de04c0bad1179b94b09eebb48c569465c","unresolved":false,"context_lines":[{"line_number":105,"context_line":"those fields. However, both of these approaches suffer from the fact"},{"line_number":106,"context_line":"that they\u0027re guessing based on circumstantial data instead of a record"},{"line_number":107,"context_line":"that tells us an evacuation was performed."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"Data model impact"},{"line_number":110,"context_line":"-----------------"},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_fc82a1ff","line":108,"in_reply_to":"3acd31a7_3c2db946","updated":"2015-04-30 21:57:07.000000000","message":"The problem with that is that we can\u0027t record multiple successive evacuations in a row before the first host gets restored. See mikal\u0027s question on an earlier revision. We could make it a list on the instance, but honestly, we have this migration object as a task record of what happened. It has the things in it we need (besides the reason). We should use it :)","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":7166,"name":"Sylvain Bauza","email":"sbauza@redhat.com","username":"sbauza"},"change_message_id":"ebb229abf921d46f84f22d5084a156f03ed0d990","unresolved":false,"context_lines":[{"line_number":105,"context_line":"those fields. However, both of these approaches suffer from the fact"},{"line_number":106,"context_line":"that they\u0027re guessing based on circumstantial data instead of a record"},{"line_number":107,"context_line":"that tells us an evacuation was performed."},{"line_number":108,"context_line":""},{"line_number":109,"context_line":"Data model impact"},{"line_number":110,"context_line":"-----------------"},{"line_number":111,"context_line":""}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_dc47e575","line":108,"in_reply_to":"3acd31a7_fc82a1ff","updated":"2015-04-30 22:13:18.000000000","message":"Agreed, as I said, your proposal is the best I can figure out. I was just opening a can of worms when speaking about alternatives, so you can blame me for that :-)","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"ada9b91ebc78182001bcbaa27a9a669f3c88377c","unresolved":false,"context_lines":[{"line_number":111,"context_line":""},{"line_number":112,"context_line":"This would re-use our existing migration data structure, so data model"},{"line_number":113,"context_line":"impact would be minor. We would need to add at least one field to"},{"line_number":114,"context_line":"track the method of the migration. This would be something like"},{"line_number":115,"context_line":"\"resize\", \"migrate\", or \"evacuate\". This will let us extend the"},{"line_number":116,"context_line":"migration log model to live migration as well as any other methods of"},{"line_number":117,"context_line":"moving instances around so that the log becomes accurate enough to"}],"source_content_type":"text/x-rst","patch_set":3,"id":"3acd31a7_6a17ea51","line":114,"range":{"start_line":114,"start_character":10,"end_line":114,"end_character":16},"updated":"2015-04-30 19:12:48.000000000","message":"Agree with you above, I changed it from reason to type (ish) here, per the feedback.","commit_id":"7bcfe2d43f56ec0e5af035c0269d71ea8bd3d627"}],"specs/liberty/robustify-evacuate.rst":[{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"8a337791faf71f09b2477e2cf8b67a22b75b8474","unresolved":false,"context_lines":[{"line_number":17,"context_line":"time. It does this by simply rebuilding the instance elsewhere. When"},{"line_number":18,"context_line":"the initial compute host comes back up, some logic on startup looks at"},{"line_number":19,"context_line":"instances that appear to have been relocated and deletes them from the"},{"line_number":20,"context_line":"local machine. This is problematic in practice, as we have very little"},{"line_number":21,"context_line":"information on which to base a decision to destroy data -- making it"},{"line_number":22,"context_line":"effectively a guess."},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a80dd14_c7b09a8d","line":20,"updated":"2015-03-13 11:56:49.000000000","message":"Is it worth spelling out why the suspected migrated instances must be deleted from the source node?\n\n(presumably to avoid data corruption if the same instance is run on two different nodes?)","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"c884807d64c771e25da469d237f2efd67f3b0f7d","unresolved":false,"context_lines":[{"line_number":17,"context_line":"time. It does this by simply rebuilding the instance elsewhere. When"},{"line_number":18,"context_line":"the initial compute host comes back up, some logic on startup looks at"},{"line_number":19,"context_line":"instances that appear to have been relocated and deletes them from the"},{"line_number":20,"context_line":"local machine. This is problematic in practice, as we have very little"},{"line_number":21,"context_line":"information on which to base a decision to destroy data -- making it"},{"line_number":22,"context_line":"effectively a guess."},{"line_number":23,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5aa745d1_c36d0ace","line":20,"in_reply_to":"9a80dd14_c7b09a8d","updated":"2015-04-15 19:19:30.000000000","message":"Technically, it\u0027s just because the local files and actual vm are no longer relevant. I suppose that if we had a volume attached there could be some corruption happening, although I think that\u0027s probably suitably protected by cinder no longer exporting the volume to the original host.\n\nAnyway, I\u0027ll add some words here to try to make it crystal clear.","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":9796,"name":"ChangBo Guo","email":"glongwave@gmail.com","username":"gcb"},"change_message_id":"e733ec21b44025a2d841b545896de98b89d4fa88","unresolved":false,"context_lines":[{"line_number":24,"context_line":"This spec proposes to improve the robustness of this process by"},{"line_number":25,"context_line":"explicitly recording the event of an instance evacuation such that the"},{"line_number":26,"context_line":"recovering host has a very clear indication that it should delete the"},{"line_number":27,"context_line":"local copy of an instance."},{"line_number":28,"context_line":""},{"line_number":29,"context_line":""},{"line_number":30,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a80dd14_446aa881","line":27,"updated":"2015-03-16 06:20:00.000000000","message":"Thanks for raising this up,  this spec looks likes focusing  on  how  to decide whether  source compute host should delete local machine or not .  For evacuate robustness, there is lots of things to handle .  For example, I have been working on bug[1],  we haven\u0027t  exception handlers for possible failures. \nso,  we need another spec or include them in this one ?\n\n [1] https://bugs.launchpad.net/nova/+bug/1404795","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"c884807d64c771e25da469d237f2efd67f3b0f7d","unresolved":false,"context_lines":[{"line_number":24,"context_line":"This spec proposes to improve the robustness of this process by"},{"line_number":25,"context_line":"explicitly recording the event of an instance evacuation such that the"},{"line_number":26,"context_line":"recovering host has a very clear indication that it should delete the"},{"line_number":27,"context_line":"local copy of an instance."},{"line_number":28,"context_line":""},{"line_number":29,"context_line":""},{"line_number":30,"context_line":"Problem description"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5aa745d1_00188883","line":27,"in_reply_to":"9a80dd14_446aa881","updated":"2015-04-15 19:19:30.000000000","message":"This spec is about a very specific improvement, as you note. Perhaps I should rename it to something like:\n\n robustify-evacuate-recovery\n\n?","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"8a337791faf71f09b2477e2cf8b67a22b75b8474","unresolved":false,"context_lines":[{"line_number":31,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a80dd14_67cd2e31","line":34,"updated":"2015-03-13 11:56:49.000000000","message":"s/reltatively/relatively/","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"c884807d64c771e25da469d237f2efd67f3b0f7d","unresolved":false,"context_lines":[{"line_number":31,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":32,"context_line":""},{"line_number":33,"context_line":"The evacuate cleanup logic in Nova guesses about whether to destroy"},{"line_number":34,"context_line":"data. It shouldn\u0027t guess, as it\u0027s reltatively easy to cause Nova to"},{"line_number":35,"context_line":"guess poorly, which results in data loss."},{"line_number":36,"context_line":""},{"line_number":37,"context_line":"Use Cases"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5aa745d1_231a6e6f","line":34,"in_reply_to":"9a80dd14_67cd2e31","updated":"2015-04-15 19:19:30.000000000","message":"egads! emacs-rst-aspell fail!","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"8a337791faf71f09b2477e2cf8b67a22b75b8474","unresolved":false,"context_lines":[{"line_number":38,"context_line":"----------"},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"A developer wants to enable evacuation on their cloud and have Nova"},{"line_number":41,"context_line":"only destroy data when it has been safelt moved to another host."},{"line_number":42,"context_line":""},{"line_number":43,"context_line":"Project Priority"},{"line_number":44,"context_line":"-----------------"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9a80dd14_c76dfa33","line":41,"updated":"2015-03-13 11:56:49.000000000","message":"s/safelt/safely/","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":8276,"name":"kiwik","email":"chenrui.momo@gmail.com","username":"kiwik"},"change_message_id":"8c545503dd57014b40fa2d0539f1e4bc30501b1e","unresolved":false,"context_lines":[{"line_number":61,"context_line":" 1. Check for unconfirmed evacuation migrations where it is the source"},{"line_number":62,"context_line":"    host"},{"line_number":63,"context_line":" 2. Delete the local data, if still present"},{"line_number":64,"context_line":" 3. Mark the migration as completed"},{"line_number":65,"context_line":""},{"line_number":66,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":67,"context_line":"current migration object so that we can keep user/admin-initiated"}],"source_content_type":"text/x-rst","patch_set":1,"id":"3ab451fb_39f9784e","line":64,"updated":"2015-04-17 02:53:47.000000000","message":"When this feature is implemented, the option \u0027destroy_after_evacuate\u0027 will be deprecated, right?","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"ad1f9d683d142a35d1a80e663ceee06119ef0ec4","unresolved":false,"context_lines":[{"line_number":61,"context_line":" 1. Check for unconfirmed evacuation migrations where it is the source"},{"line_number":62,"context_line":"    host"},{"line_number":63,"context_line":" 2. Delete the local data, if still present"},{"line_number":64,"context_line":" 3. Mark the migration as completed"},{"line_number":65,"context_line":""},{"line_number":66,"context_line":"This means that we need to add a \"type\" or \"reason\" field to the"},{"line_number":67,"context_line":"current migration object so that we can keep user/admin-initiated"}],"source_content_type":"text/x-rst","patch_set":1,"id":"dac4157f_cf41530d","line":64,"in_reply_to":"3ab451fb_39f9784e","updated":"2015-04-21 14:12:39.000000000","message":"Yes, added this to the next revision.","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"03b69b4bf9cb0ee8bd837a5a1538bf5e25a26b6a","unresolved":false,"context_lines":[{"line_number":166,"context_line":"Testing"},{"line_number":167,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Existing tempest and functional tests will ensure that we don\u0027t break"},{"line_number":170,"context_line":"the evacuation path. Testing of nova-compute startup is tricky, but"},{"line_number":171,"context_line":"should be easily covered by unit tests of the evacuation cleanup logic."},{"line_number":172,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aaa499b_7a220003","line":169,"updated":"2015-04-14 15:20:38.000000000","message":"We don\u0027t have any tempest testing of the os-evacuate API today, probably because we didn\u0027t have multi-node testing in the gate until recently with the aiopcpu job.\n\nSo we should add an evacuate test to Tempest as a pre-req for this and run it on the aiopcpu job.  I can start the paperwork for Tempest.","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"9d07e504b4af397a430ae99047622472645aed32","unresolved":false,"context_lines":[{"line_number":166,"context_line":"Testing"},{"line_number":167,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Existing tempest and functional tests will ensure that we don\u0027t break"},{"line_number":170,"context_line":"the evacuation path. Testing of nova-compute startup is tricky, but"},{"line_number":171,"context_line":"should be easily covered by unit tests of the evacuation cleanup logic."},{"line_number":172,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5aa745d1_2ce25ba0","line":169,"in_reply_to":"5aa745d1_e3780692","updated":"2015-04-15 21:04:03.000000000","message":"The problem is that tempest is running all of the tests concurrently, so we\u0027d basically need a lock fixture on the evacuate tests to not run any other tests that require the compute API at the same time, which I don\u0027t think exists today.  idk, we can bug the tempest peeps about how feasible that is.","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":6873,"name":"Matt Riedemann","email":"mriedem.os@gmail.com","username":"mriedem"},"change_message_id":"76d961e96ce1011b02a235aa1cea7a52b0b73a9e","unresolved":false,"context_lines":[{"line_number":166,"context_line":"Testing"},{"line_number":167,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Existing tempest and functional tests will ensure that we don\u0027t break"},{"line_number":170,"context_line":"the evacuation path. Testing of nova-compute startup is tricky, but"},{"line_number":171,"context_line":"should be easily covered by unit tests of the evacuation cleanup logic."},{"line_number":172,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"7aaa499b_f0506f0c","line":169,"in_reply_to":"7aaa499b_7a220003","updated":"2015-04-14 15:49:19.000000000","message":"Turns out we might not be able to test this in Tempest since the compute API checks that the compute service is down before casting to conductor:\n\nhttp://git.openstack.org/cgit/openstack/nova/tree/nova/compute/api.py#n3254\n\nWhich wouldn\u0027t be the case in Tempest, unless we wrote some code to do that, e.g. lock out the evacuate tests in Tempest from running concurrently with any other tests, bring down the compute service on A, do the evacuate to B, then bring up the service on A in tearDown.","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"},{"author":{"_account_id":4393,"name":"Dan Smith","email":"dms@danplanet.com","username":"danms"},"change_message_id":"c884807d64c771e25da469d237f2efd67f3b0f7d","unresolved":false,"context_lines":[{"line_number":166,"context_line":"Testing"},{"line_number":167,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":168,"context_line":""},{"line_number":169,"context_line":"Existing tempest and functional tests will ensure that we don\u0027t break"},{"line_number":170,"context_line":"the evacuation path. Testing of nova-compute startup is tricky, but"},{"line_number":171,"context_line":"should be easily covered by unit tests of the evacuation cleanup logic."},{"line_number":172,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5aa745d1_e3780692","line":169,"in_reply_to":"7aaa499b_f0506f0c","updated":"2015-04-15 19:19:30.000000000","message":"We can down the service from the API. Can\u0027t we use that to force the behavior we want?","commit_id":"5a878c9f70d9a34790f4b5b17d01560ea1b8b52c"}]}
