| *** haomaiwang has quit IRC | 00:01 | |
| *** haomaiwang has joined #openstack-cinder | 00:01 | |
| *** lprice has quit IRC | 00:03 | |
| *** jamielennox is now known as jamielennox|away | 00:04 | |
| *** salv-orlando has quit IRC | 00:06 | |
| *** jamielennox|away is now known as jamielennox | 00:07 | |
| *** p0rtal has joined #openstack-cinder | 00:13 | |
| *** stevemar_ has joined #openstack-cinder | 00:16 | |
| *** p0rtal has quit IRC | 00:17 | |
| *** stevemar_ has quit IRC | 00:20 | |
| *** apoorvad has quit IRC | 00:23 | |
| *** EinstCrazy has quit IRC | 00:25 | |
| *** takedakn has quit IRC | 00:27 | |
| *** takedakn has joined #openstack-cinder | 00:28 | |
| *** takedakn has quit IRC | 00:32 | |
| *** jerrygb has joined #openstack-cinder | 00:34 | |
| *** jerrygb has quit IRC | 00:39 | |
| *** jamielennox is now known as jamielennox|away | 00:45 | |
| openstackgerrit | Thang Pham proposed openstack/cinder: WIP: Backwards compatible update to create_volume https://review.openstack.org/240735 | 00:48 |
|---|---|---|
| *** takedakn has joined #openstack-cinder | 00:49 | |
| *** apoorvad has joined #openstack-cinder | 00:52 | |
| *** rhagarty has quit IRC | 00:54 | |
| *** apoorvad has quit IRC | 00:56 | |
| *** apoorvad has joined #openstack-cinder | 00:56 | |
| *** jamielennox|away is now known as jamielennox | 00:57 | |
| *** chlong has quit IRC | 00:59 | |
| *** haomaiwang has quit IRC | 01:01 | |
| *** haomaiwang has joined #openstack-cinder | 01:01 | |
| *** jamielennox is now known as jamielennox|away | 01:04 | |
| *** salv-orlando has joined #openstack-cinder | 01:07 | |
| *** lixiaoy1 has joined #openstack-cinder | 01:09 | |
| *** jerrygb has joined #openstack-cinder | 01:10 | |
| *** takedakn has quit IRC | 01:11 | |
| *** apoorvad has quit IRC | 01:12 | |
| *** chlong has joined #openstack-cinder | 01:12 | |
| *** EinstCrazy has joined #openstack-cinder | 01:13 | |
| *** jamielennox|away is now known as jamielennox | 01:13 | |
| *** jerrygb has quit IRC | 01:15 | |
| *** yangxi has joined #openstack-cinder | 01:17 | |
| *** zhenguo has joined #openstack-cinder | 01:20 | |
| *** hemna has joined #openstack-cinder | 01:27 | |
| *** haomaiwang has quit IRC | 01:28 | |
| *** Lee1092 has joined #openstack-cinder | 01:31 | |
| *** chenying has joined #openstack-cinder | 01:31 | |
| *** terryyao has joined #openstack-cinder | 01:36 | |
| openstackgerrit | chenying proposed openstack/cinder: Fix can not adding a new zone with Cisco FC zone driver https://review.openstack.org/240649 | 01:36 |
| *** chenying has quit IRC | 01:36 | |
| *** smatzek has joined #openstack-cinder | 01:47 | |
| *** stevemar_ has joined #openstack-cinder | 01:52 | |
| *** rhagarty has joined #openstack-cinder | 01:54 | |
| *** sileht has quit IRC | 01:54 | |
| *** stevemar_ has quit IRC | 01:56 | |
| *** yangxi has quit IRC | 01:59 | |
| *** yangxi has joined #openstack-cinder | 02:03 | |
| *** p0rtal has joined #openstack-cinder | 02:05 | |
| *** haomaiwang has joined #openstack-cinder | 02:10 | |
| *** p0rtal has quit IRC | 02:10 | |
| *** jerrygb has joined #openstack-cinder | 02:18 | |
| *** yangxi has quit IRC | 02:19 | |
| *** jerrygb has quit IRC | 02:23 | |
| *** yangxi has joined #openstack-cinder | 02:23 | |
| *** jamielennox is now known as jamielennox|away | 02:23 | |
| *** hemna has quit IRC | 02:24 | |
| *** hemna has joined #openstack-cinder | 02:25 | |
| *** salv-orlando has quit IRC | 02:28 | |
| *** hemna has quit IRC | 02:29 | |
| *** jamielennox|away is now known as jamielennox | 02:31 | |
| openstackgerrit | XinXiaohui proposed openstack/cinder: Calculate virtual free capacity and notify https://review.openstack.org/206923 | 02:32 |
| *** rex_lee_ has quit IRC | 02:35 | |
| *** smatzek has quit IRC | 02:36 | |
| *** haomaiwang has quit IRC | 03:01 | |
| *** 7F1AAWWJJ has joined #openstack-cinder | 03:01 | |
| *** terryyao has quit IRC | 03:15 | |
| *** jerrygb has joined #openstack-cinder | 03:19 | |
| *** salv-orlando has joined #openstack-cinder | 03:27 | |
| *** daneyon has joined #openstack-cinder | 03:35 | |
| *** jerrygb has quit IRC | 03:38 | |
| *** daneyon has quit IRC | 03:41 | |
| openstackgerrit | Cedric Zhuang proposed openstack/cinder: Add retype logic in manage_existing for VNX https://review.openstack.org/209176 | 03:42 |
| *** hemna has joined #openstack-cinder | 03:46 | |
| *** dongc has joined #openstack-cinder | 03:47 | |
| *** gouthamr has quit IRC | 03:54 | |
| *** yangxi has quit IRC | 03:59 | |
| *** 7F1AAWWJJ has quit IRC | 04:01 | |
| *** haomaiwang has joined #openstack-cinder | 04:01 | |
| *** yangxi has joined #openstack-cinder | 04:01 | |
| *** yangxi has quit IRC | 04:03 | |
| *** hideme_ has quit IRC | 04:11 | |
| *** p0rtal has joined #openstack-cinder | 04:11 | |
| *** hideme_ has joined #openstack-cinder | 04:11 | |
| *** dims has quit IRC | 04:13 | |
| *** jerrygb has joined #openstack-cinder | 04:15 | |
| *** links has joined #openstack-cinder | 04:15 | |
| *** lixiaoy1 has quit IRC | 04:15 | |
| *** hemna has quit IRC | 04:38 | |
| *** yangxi has joined #openstack-cinder | 04:38 | |
| *** jerrygb has quit IRC | 04:39 | |
| *** p0rtal has quit IRC | 04:57 | |
| *** p0rtal has joined #openstack-cinder | 04:58 | |
| *** haomaiwang has quit IRC | 05:01 | |
| *** haomaiwang has joined #openstack-cinder | 05:01 | |
| *** salv-orlando has quit IRC | 05:02 | |
| *** p0rtal has quit IRC | 05:02 | |
| *** lixiaoy1 has joined #openstack-cinder | 05:06 | |
| openstackgerrit | xing-yang proposed openstack/cinder-specs: Integrate replication with consistency group https://review.openstack.org/229722 | 05:09 |
| *** lixiaoy11 has joined #openstack-cinder | 05:15 | |
| *** lixiaoy1 has quit IRC | 05:17 | |
| *** deepakcs has joined #openstack-cinder | 05:18 | |
| *** sgotliv has quit IRC | 05:19 | |
| *** sgotliv has joined #openstack-cinder | 05:19 | |
| *** lixiaoy1 has joined #openstack-cinder | 05:22 | |
| *** lixiaoy11 has quit IRC | 05:24 | |
| *** lixiaoy11 has joined #openstack-cinder | 05:25 | |
| *** lixiaoy1 has quit IRC | 05:27 | |
| *** boris-42 has joined #openstack-cinder | 05:28 | |
| *** salv-orlando has joined #openstack-cinder | 05:40 | |
| *** yangxi has quit IRC | 05:41 | |
| *** shausy has joined #openstack-cinder | 05:43 | |
| *** zenpac has quit IRC | 05:48 | |
| *** yangxi has joined #openstack-cinder | 05:55 | |
| *** dongc has quit IRC | 05:55 | |
| *** haomaiwang has quit IRC | 06:01 | |
| *** haomaiwang has joined #openstack-cinder | 06:01 | |
| *** lprice has joined #openstack-cinder | 06:04 | |
| *** jamielennox is now known as jamielennox|away | 06:06 | |
| *** links has quit IRC | 06:07 | |
| *** sileht has joined #openstack-cinder | 06:07 | |
| *** salv-orlando has quit IRC | 06:20 | |
| *** cfriesen_ has quit IRC | 06:22 | |
| *** sgotliv has quit IRC | 06:26 | |
| *** zerda has joined #openstack-cinder | 06:32 | |
| *** takedakn has joined #openstack-cinder | 06:41 | |
| *** takedakn has quit IRC | 06:41 | |
| *** links has joined #openstack-cinder | 06:44 | |
| *** zhangjn has joined #openstack-cinder | 06:45 | |
| *** chlong has quit IRC | 06:48 | |
| *** nkrinner has joined #openstack-cinder | 07:00 | |
| *** jamielennox|away is now known as jamielennox | 07:01 | |
| *** haomaiwang has quit IRC | 07:01 | |
| *** 7F1AAWXGU has joined #openstack-cinder | 07:01 | |
| *** jith_ has joined #openstack-cinder | 07:12 | |
| jith_ | hi all.. if i configure ceph as backend for cinder... is it possible to use the same ceph backend for other purpose like swift?? or cinder should have a dedicated ceph storage? | 07:14 |
| nikeshm | jith_: i think you can use same ceph for both ceph volume driver and ceph backup driver | 07:26 |
| *** salv-orlando has joined #openstack-cinder | 07:30 | |
| nikeshm | jith_ : https://github.com/ceph/ceph-devstack/blob/master/localrc you can use this for single node ceph devstack and play with ceph volume and backup driver | 07:34 |
| nikeshm | modify it little bit | 07:34 |
| jith_ | nikeshm: thanks | 07:34 |
| *** salv-orlando has quit IRC | 07:35 | |
| *** kmartin has quit IRC | 07:35 | |
| *** flaper87 has quit IRC | 07:35 | |
| *** kmartin has joined #openstack-cinder | 07:36 | |
| *** NightKhaos has quit IRC | 07:36 | |
| *** links has quit IRC | 07:37 | |
| *** flaper87 has joined #openstack-cinder | 07:37 | |
| *** NightKhaos has joined #openstack-cinder | 07:37 | |
| *** links has joined #openstack-cinder | 07:39 | |
| jith_ | nikeshm: i am a newbie in ceph.. ceph volume driver and backup driver for cinder volumes??? because i saw a reference like https://github.com/openstack/swift-ceph-backend for configuring swift.. i dont know about the internal operations in ceph | 07:40 |
| *** jwcroppe has joined #openstack-cinder | 07:45 | |
| *** anshul has joined #openstack-cinder | 07:46 | |
| *** salv-orlando has joined #openstack-cinder | 07:54 | |
| *** alexschm has joined #openstack-cinder | 07:58 | |
| *** 7F1AAWXGU has quit IRC | 08:01 | |
| *** haomaiwang has joined #openstack-cinder | 08:01 | |
| *** shausy has quit IRC | 08:13 | |
| *** shausy has joined #openstack-cinder | 08:14 | |
| *** haomaiwang has quit IRC | 08:14 | |
| *** haomaiwang has joined #openstack-cinder | 08:14 | |
| *** haomaiwang has quit IRC | 08:15 | |
| *** haomaiwa_ has joined #openstack-cinder | 08:15 | |
| *** bapalm has quit IRC | 08:19 | |
| *** markus_z has joined #openstack-cinder | 08:20 | |
| *** bapalm has joined #openstack-cinder | 08:24 | |
| *** apoorvad has joined #openstack-cinder | 08:24 | |
| *** chenying1 has joined #openstack-cinder | 08:25 | |
| *** stevemar_ has joined #openstack-cinder | 08:25 | |
| *** yrabl has joined #openstack-cinder | 08:26 | |
| *** aarefiev has joined #openstack-cinder | 08:29 | |
| *** salv-orlando has quit IRC | 08:30 | |
| *** yrabl has quit IRC | 08:33 | |
| *** yrabl has joined #openstack-cinder | 08:34 | |
| openstackgerrit | Hao Li proposed openstack/cinder: return backup parent_id when list/show backups https://review.openstack.org/239364 | 08:36 |
| *** vignesh1 has joined #openstack-cinder | 08:38 | |
| *** ZZelle has joined #openstack-cinder | 08:39 | |
| *** salv-orlando has joined #openstack-cinder | 08:40 | |
| *** sgotliv has joined #openstack-cinder | 08:46 | |
| *** bluex has joined #openstack-cinder | 08:48 | |
| *** bluex has quit IRC | 08:48 | |
| *** bluex has joined #openstack-cinder | 08:49 | |
| *** haomaiwa_ has quit IRC | 09:01 | |
| *** haomaiwang has joined #openstack-cinder | 09:01 | |
| nikeshm | jiith_ : http://docs.openstack.org/liberty/config-reference/content/ceph-backup-driver.html | 09:09 |
| nikeshm | http://docs.openstack.org/liberty/config-reference/content/ceph-rados.html | 09:10 |
| *** jordanP has joined #openstack-cinder | 09:10 | |
| nikeshm | jiith_: this should give most of thing https://github.com/ceph/ceph-devstack/blob/master/localrc | 09:11 |
| nikeshm | may be you have to change cinder.conf for enabling ceph backup driver | 09:11 |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/240797 | 09:12 |
| *** manas has joined #openstack-cinder | 09:14 | |
| *** chenying1 has quit IRC | 09:16 | |
| *** chenying1 has joined #openstack-cinder | 09:17 | |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 09:17 |
| *** manas has quit IRC | 09:17 | |
| *** apoorvad has quit IRC | 09:28 | |
| *** jistr has joined #openstack-cinder | 09:28 | |
| *** stevemar_ has quit IRC | 09:46 | |
| *** lprice has quit IRC | 09:48 | |
| *** lixiaoy11 has quit IRC | 09:53 | |
| *** vgridnev has joined #openstack-cinder | 09:57 | |
| *** zhenguo has quit IRC | 09:57 | |
| *** haomaiwang has quit IRC | 10:01 | |
| *** haomaiwang has joined #openstack-cinder | 10:01 | |
| *** yuriy_n17 has joined #openstack-cinder | 10:03 | |
| *** yangxi has quit IRC | 10:14 | |
| *** jaypipes has joined #openstack-cinder | 10:15 | |
| *** yangxi has joined #openstack-cinder | 10:16 | |
| *** deepakcs has quit IRC | 10:19 | |
| *** deepakcs has joined #openstack-cinder | 10:19 | |
| *** yangxi has quit IRC | 10:22 | |
| openstackgerrit | Thang Pham proposed openstack/cinder: Update create_volume API to use versionedobjects https://review.openstack.org/240735 | 10:27 |
| jith_ | nikeshm: thanks.. i got it | 10:32 |
| *** jith_ has quit IRC | 10:34 | |
| openstackgerrit | zhangsong proposed openstack/cinder: SheepdogDriver: Improve get_volume_stats operation https://review.openstack.org/240820 | 10:35 |
| *** salv-orlando has quit IRC | 10:36 | |
| *** salv-orlando has joined #openstack-cinder | 10:36 | |
| *** jwcroppe has quit IRC | 10:37 | |
| *** jwcroppe has joined #openstack-cinder | 10:38 | |
| *** haomaiwang has quit IRC | 10:38 | |
| *** haomaiwang has joined #openstack-cinder | 10:39 | |
| *** zhangjn has quit IRC | 10:42 | |
| *** jwcroppe has quit IRC | 10:43 | |
| *** haomaiwang has quit IRC | 10:43 | |
| *** EinstCrazy has quit IRC | 10:45 | |
| *** haomaiwang has joined #openstack-cinder | 10:45 | |
| *** salv-orlando has quit IRC | 10:47 | |
| *** salv-orlando has joined #openstack-cinder | 10:48 | |
| *** subscope has joined #openstack-cinder | 10:48 | |
| *** haomaiwang has quit IRC | 11:01 | |
| *** haomaiwang has joined #openstack-cinder | 11:01 | |
| *** links has quit IRC | 11:03 | |
| *** yusuke has quit IRC | 11:07 | |
| *** links has joined #openstack-cinder | 11:09 | |
| *** dims has joined #openstack-cinder | 11:12 | |
| *** EinstCrazy has joined #openstack-cinder | 11:14 | |
| *** Zhongjun has joined #openstack-cinder | 11:18 | |
| openstackgerrit | Wilson Liu proposed openstack/cinder: Huawei: add manage/unmanage volume support https://review.openstack.org/240413 | 11:21 |
| *** dims has quit IRC | 11:21 | |
| *** salv-orlando has quit IRC | 11:23 | |
| *** haomaiwang has quit IRC | 11:24 | |
| *** salv-orlando has joined #openstack-cinder | 11:25 | |
| *** smatzek has joined #openstack-cinder | 11:32 | |
| *** cdelatte has joined #openstack-cinder | 11:33 | |
| *** geguileo is now known as geguileo_onPTO | 11:35 | |
| *** flaper87 has quit IRC | 11:41 | |
| *** flaper87 has joined #openstack-cinder | 11:41 | |
| *** dave-mccowan has joined #openstack-cinder | 11:47 | |
| *** zhangjn has joined #openstack-cinder | 11:47 | |
| *** zhangjn has quit IRC | 11:47 | |
| *** zhangjn has joined #openstack-cinder | 11:48 | |
| *** stevemar_ has joined #openstack-cinder | 11:58 | |
| *** subscope has quit IRC | 11:59 | |
| *** subscope has joined #openstack-cinder | 12:00 | |
| *** subscope has quit IRC | 12:02 | |
| *** subscope has joined #openstack-cinder | 12:02 | |
| *** stevemar_ has quit IRC | 12:02 | |
| *** stevemar_ has joined #openstack-cinder | 12:11 | |
| *** delattec has joined #openstack-cinder | 12:22 | |
| *** cdelatte has quit IRC | 12:22 | |
| *** ondergetekende has quit IRC | 12:31 | |
| *** onder has joined #openstack-cinder | 12:32 | |
| *** deepakcs has quit IRC | 12:40 | |
| *** zao has joined #openstack-cinder | 12:40 | |
| *** kevincarr1991 has joined #openstack-cinder | 12:54 | |
| *** porrua has joined #openstack-cinder | 12:56 | |
| *** salv-orlando has quit IRC | 12:58 | |
| *** subscope has quit IRC | 12:59 | |
| *** jerrygb has joined #openstack-cinder | 13:13 | |
| *** zhangjn has quit IRC | 13:16 | |
| *** jerrygb has quit IRC | 13:18 | |
| openstackgerrit | Szymon Wróblewski proposed openstack/cinder: Add some missing fields to Volume object https://review.openstack.org/240847 | 13:18 |
| *** zhangjn has joined #openstack-cinder | 13:19 | |
| *** jerrygb has joined #openstack-cinder | 13:20 | |
| *** edmondsw has joined #openstack-cinder | 13:28 | |
| *** subscope has joined #openstack-cinder | 13:29 | |
| *** dims has joined #openstack-cinder | 13:33 | |
| *** rzerda has joined #openstack-cinder | 13:36 | |
| *** timcl has joined #openstack-cinder | 13:37 | |
| *** Zhongjun has quit IRC | 13:37 | |
| *** zerda has quit IRC | 13:38 | |
| *** sgotliv has quit IRC | 13:38 | |
| *** dims has quit IRC | 13:39 | |
| *** bswartz has quit IRC | 13:39 | |
| *** diablo_rojo has joined #openstack-cinder | 13:39 | |
| *** krtaylor has quit IRC | 13:43 | |
| *** stevemar_ has quit IRC | 13:43 | |
| *** salv-orlando has joined #openstack-cinder | 13:45 | |
| *** jerrygb has quit IRC | 13:45 | |
| *** jerrygb has joined #openstack-cinder | 13:46 | |
| *** ociuhandu has joined #openstack-cinder | 13:51 | |
| *** jerrygb has quit IRC | 13:51 | |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 13:52 |
| *** sgotliv has joined #openstack-cinder | 13:54 | |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 13:56 |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 13:58 |
| *** gouthamr has joined #openstack-cinder | 13:59 | |
| *** links has quit IRC | 14:00 | |
| *** mc_nair has joined #openstack-cinder | 14:01 | |
| *** akerr has joined #openstack-cinder | 14:05 | |
| *** akerr_ has joined #openstack-cinder | 14:06 | |
| *** xyang1 has joined #openstack-cinder | 14:08 | |
| *** cfriesen_ has joined #openstack-cinder | 14:10 | |
| *** dims has joined #openstack-cinder | 14:10 | |
| *** xyang1 has quit IRC | 14:10 | |
| *** yangxi has joined #openstack-cinder | 14:10 | |
| *** akerr has quit IRC | 14:10 | |
| *** xyang1 has joined #openstack-cinder | 14:10 | |
| *** sileht has quit IRC | 14:14 | |
| *** dustins has joined #openstack-cinder | 14:14 | |
| *** Yogi1 has joined #openstack-cinder | 14:16 | |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 14:17 |
| *** breitz has joined #openstack-cinder | 14:19 | |
| *** yangxi has quit IRC | 14:20 | |
| *** jgregor has joined #openstack-cinder | 14:21 | |
| *** alejandrito has joined #openstack-cinder | 14:22 | |
| *** julim has joined #openstack-cinder | 14:22 | |
| *** julim has quit IRC | 14:23 | |
| *** rzerda has quit IRC | 14:24 | |
| *** subscope has quit IRC | 14:26 | |
| *** vincent_hou has joined #openstack-cinder | 14:27 | |
| *** hideme_ has quit IRC | 14:27 | |
| *** jerrygb has joined #openstack-cinder | 14:28 | |
| *** hideme_ has joined #openstack-cinder | 14:28 | |
| openstackgerrit | Szymon Wróblewski proposed openstack/cinder: Add some missing fields to Volume object https://review.openstack.org/240847 | 14:28 |
| *** mriedem has joined #openstack-cinder | 14:29 | |
| *** scottda has joined #openstack-cinder | 14:32 | |
| *** bswartz has joined #openstack-cinder | 14:32 | |
| *** zz_john5223 is now known as john5223 | 14:32 | |
| vincent_hou | jbernard: hi | 14:36 |
| jbernard | vincent_hou: yi | 14:36 |
| jbernard | vincent_hou: hi | 14:36 |
| vincent_hou | jbernard: how are you? | 14:36 |
| *** zerda has joined #openstack-cinder | 14:37 | |
| jbernard | vincent_hou: im doing great, you? | 14:37 |
| vincent_hou | jbernard: I am fine too. | 14:37 |
| *** markvoelker has joined #openstack-cinder | 14:40 | |
| *** zerda has quit IRC | 14:41 | |
| *** subscope has joined #openstack-cinder | 14:41 | |
| *** Trident has joined #openstack-cinder | 14:48 | |
| *** subscope has quit IRC | 14:54 | |
| *** zao has left #openstack-cinder | 14:55 | |
| *** willsama has joined #openstack-cinder | 14:55 | |
| *** subscope has joined #openstack-cinder | 14:55 | |
| *** subscope has quit IRC | 14:57 | |
| *** dustins has quit IRC | 14:59 | |
| *** yrabl has quit IRC | 15:01 | |
| *** ociuhandu has quit IRC | 15:01 | |
| *** ociuhandu has joined #openstack-cinder | 15:01 | |
| openstackgerrit | Thang Pham proposed openstack/cinder: Update get/delete_volume API to use versionedobjects https://review.openstack.org/177054 | 15:03 |
| *** baumann has joined #openstack-cinder | 15:03 | |
| *** krtaylor has joined #openstack-cinder | 15:06 | |
| *** jerrygb has quit IRC | 15:06 | |
| *** jerrygb has joined #openstack-cinder | 15:06 | |
| *** haomaiwang has joined #openstack-cinder | 15:07 | |
| *** jerrygb has quit IRC | 15:07 | |
| *** jerrygb has joined #openstack-cinder | 15:07 | |
| *** zhangjn has quit IRC | 15:08 | |
| *** aix has quit IRC | 15:09 | |
| *** subscope has joined #openstack-cinder | 15:09 | |
| *** delattec has quit IRC | 15:10 | |
| *** delattec has joined #openstack-cinder | 15:11 | |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 15:13 |
| openstackgerrit | Hao Li proposed openstack/cinder-specs: List backups return "parent_id" field https://review.openstack.org/239674 | 15:14 |
| openstackgerrit | abhiram moturi proposed openstack/cinder: Zfssaiscsi driver should not use 'default' initiator group https://review.openstack.org/238635 | 15:14 |
| *** anshul has quit IRC | 15:14 | |
| *** vincent_hou has quit IRC | 15:20 | |
| *** ntpttr_ has quit IRC | 15:20 | |
| *** vgridnev has quit IRC | 15:24 | |
| *** cvstealt1 has joined #openstack-cinder | 15:27 | |
| *** pckizer_ has joined #openstack-cinder | 15:27 | |
| *** jgregor1 has joined #openstack-cinder | 15:28 | |
| *** sage has joined #openstack-cinder | 15:31 | |
| *** kragniz_ has joined #openstack-cinder | 15:32 | |
| *** kmartin_ has joined #openstack-cinder | 15:34 | |
| *** leseb_ has joined #openstack-cinder | 15:34 | |
| *** jgregor has quit IRC | 15:35 | |
| *** onder has quit IRC | 15:35 | |
| *** jaypipes has quit IRC | 15:35 | |
| *** flaper87 has quit IRC | 15:35 | |
| *** kmartin has quit IRC | 15:35 | |
| *** nkrinner has quit IRC | 15:35 | |
| *** geguileo_onPTO has quit IRC | 15:35 | |
| *** zul has quit IRC | 15:35 | |
| *** DuncanT has quit IRC | 15:35 | |
| *** briancurtin has quit IRC | 15:35 | |
| *** toabctl has quit IRC | 15:35 | |
| *** liewegas has quit IRC | 15:35 | |
| *** dhellmann has quit IRC | 15:35 | |
| *** jbernard has quit IRC | 15:35 | |
| *** leseb- has quit IRC | 15:35 | |
| *** pckizer has quit IRC | 15:35 | |
| *** kragniz has quit IRC | 15:35 | |
| *** cvstealth has quit IRC | 15:35 | |
| *** lyarwood has quit IRC | 15:35 | |
| *** git-harry has quit IRC | 15:35 | |
| *** dhellmann has joined #openstack-cinder | 15:36 | |
| *** toabctl has joined #openstack-cinder | 15:36 | |
| *** flaper87 has joined #openstack-cinder | 15:36 | |
| *** flaper87 has quit IRC | 15:36 | |
| *** flaper87 has joined #openstack-cinder | 15:36 | |
| *** jaypipes has joined #openstack-cinder | 15:38 | |
| *** haomaiwang has quit IRC | 15:38 | |
| *** p0rtal has joined #openstack-cinder | 15:41 | |
| *** nkrinner has joined #openstack-cinder | 15:41 | |
| *** geguileo_onPTO has joined #openstack-cinder | 15:41 | |
| *** bluex has quit IRC | 15:42 | |
| *** stevemar_ has joined #openstack-cinder | 15:43 | |
| *** briancurtin has joined #openstack-cinder | 15:44 | |
| *** p0rtal has quit IRC | 15:45 | |
| *** zul has joined #openstack-cinder | 15:47 | |
| *** kevincar_ has joined #openstack-cinder | 15:48 | |
| *** stevemar_ has quit IRC | 15:49 | |
| *** DuncanT has joined #openstack-cinder | 15:49 | |
| *** breitz has quit IRC | 15:50 | |
| *** Swanson has quit IRC | 15:50 | |
| *** jbernard has joined #openstack-cinder | 15:50 | |
| *** kevincarr1991 has quit IRC | 15:51 | |
| *** Swanson has joined #openstack-cinder | 15:53 | |
| *** breitz has joined #openstack-cinder | 15:53 | |
| *** cebruns has joined #openstack-cinder | 15:56 | |
| *** jgregor1 has quit IRC | 15:56 | |
| *** lprice has joined #openstack-cinder | 15:57 | |
| *** jungleboyj has joined #openstack-cinder | 15:57 | |
| *** baumann has quit IRC | 15:58 | |
| *** jgregor has joined #openstack-cinder | 15:58 | |
| *** jwcroppe has joined #openstack-cinder | 16:03 | |
| *** martyturner has joined #openstack-cinder | 16:03 | |
| *** baumann has joined #openstack-cinder | 16:04 | |
| *** shausy has quit IRC | 16:05 | |
| *** alexschm has quit IRC | 16:06 | |
| *** vignesh has joined #openstack-cinder | 16:07 | |
| *** onder has joined #openstack-cinder | 16:08 | |
| *** delatte has joined #openstack-cinder | 16:08 | |
| *** baumann1 has joined #openstack-cinder | 16:08 | |
| *** baumann has quit IRC | 16:08 | |
| *** dustins has joined #openstack-cinder | 16:08 | |
| *** edtubill has joined #openstack-cinder | 16:08 | |
| *** vignesh1 has quit IRC | 16:09 | |
| *** delattec has quit IRC | 16:09 | |
| *** merooney has joined #openstack-cinder | 16:11 | |
| *** mtanino has joined #openstack-cinder | 16:16 | |
| *** sgotliv has quit IRC | 16:17 | |
| *** vgridnev has joined #openstack-cinder | 16:19 | |
| *** subscope has quit IRC | 16:20 | |
| *** kevincar_ has quit IRC | 16:24 | |
| *** leeantho has joined #openstack-cinder | 16:26 | |
| *** hemnafk is now known as hemna | 16:28 | |
| hemna | morning | 16:28 |
| hemna | it's like Monday and stuff or something. | 16:29 |
| tbarron | yeah, I heard it was Monday too. Doesn't seem right though. | 16:33 |
| *** kevincarr1991 has joined #openstack-cinder | 16:34 | |
| *** yuriy_n17 has quit IRC | 16:34 | |
| diablo_rojo | tbarron: hemna yeah yeah yeah :P | 16:35 |
| tbarron | diablo_rojo: :P | 16:35 |
| diablo_rojo | tbarron: We totally trashed Jay's office while he was gone. | 16:36 |
| hemna | picts or it didn't happen. | 16:36 |
| tbarron | hemna: +1 | 16:36 |
| *** pckizer_ is now known as pckizer | 16:37 | |
| diablo_rojo | hemna: I just fb messaged you a picture. | 16:37 |
| *** sileht has joined #openstack-cinder | 16:37 | |
| diablo_rojo | hemna: tbarron Also we screwed with this mouse.. | 16:38 |
| *** jungleboyj has quit IRC | 16:38 | |
| diablo_rojo | hemna: tbarron popped the ball out and put a troll face covering the sensor. And turned it off..And put another troll face in with the batteries.. | 16:39 |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/240931 | 16:39 |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/240932 | 16:40 |
| diablo_rojo | tbarron: Here you go https://dl.pushbulletusercontent.com/JljoC6HOqCEIdXFRrLoMwDt5iKBNymWn/20151030_110640.jpg | 16:42 |
| *** stevemar_ has joined #openstack-cinder | 16:45 | |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/223803 | 16:46 |
| *** diogogmt has joined #openstack-cinder | 16:46 | |
| tbarron | diablo_rojo: nice! you guys should "fix" the storwize driver next. | 16:47 |
| *** stevemar_ has quit IRC | 16:49 | |
| *** jungleboyj has joined #openstack-cinder | 16:50 | |
| diablo_rojo | tbarron: jgregor is working on it with baumann1 | 16:54 |
| *** sileht has quit IRC | 16:55 | |
| jgriffith | mc_nair: hey... thanks for the feedback BTW | 16:55 |
| smcginnis | diablo_rojo: Nice! | 16:55 |
| openstackgerrit | xing-yang proposed openstack/cinder: Implement update_migrated_volume for the ScaleIO driver https://review.openstack.org/240937 | 16:55 |
| jgregor | :) | 16:55 |
| *** kevincarr1991 has quit IRC | 16:56 | |
| smcginnis | mc_nair: Did you still have the question about LOG.warning? | 16:56 |
| *** kevincarr1991 has joined #openstack-cinder | 16:56 | |
| *** sileht has joined #openstack-cinder | 16:57 | |
| mc_nair | jgriffith: sure thing. Thanks for the responses on those - a few of the things were as much me trying to see why you did something a certain way so I could learn. | 16:57 |
| mc_nair | smatzek: yea, that | 16:57 |
| mc_nair | ^ sorry - shaky hands + keyboard | 16:58 |
| jgriffith | mc_nair: yeah, I appreciate the questions. | 16:58 |
| jgriffith | I'll turn some of those around here later today | 16:58 |
| jgriffith | good to make me think :) | 16:58 |
| mc_nair | smcginnis: sure, that'd be great if you could explain the benefit of doing "LOG.warning(_LW("some message"))" vs. "LOG.warning(_("some message"))" | 16:59 |
| smcginnis | mc_nair: So the correct way should be LOG.warning(_LW('xxx')) | 16:59 |
| *** nkrinner has quit IRC | 17:00 | |
| smcginnis | mc_nair: There shouldn't be any doing LOG.warning(_('xxx')). If there are, then that should be fixed. | 17:00 |
| smcginnis | mc_nair: From what I understand, _Lx categorizes the strings so the folks working on translations know the level being used and prioritize accordingly. | 17:01 |
| *** garthb has joined #openstack-cinder | 17:02 | |
| smcginnis | mc_nair: The only time you wouldn't use _LW for a LOG.warning is if you assign to a string, then use that string both for logging and for setting an exception message. | 17:03 |
| smcginnis | mc_nair: A little more detail here: http://docs.openstack.org/developer/oslo.i18n/guidelines.html | 17:03 |
| *** daneyon has joined #openstack-cinder | 17:03 | |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/223803 | 17:03 |
| *** markus_z has quit IRC | 17:11 | |
| openstackgerrit | Nate Potter proposed openstack/cinder: Remove 'refresh' parameter from driver get_stats https://review.openstack.org/233328 | 17:11 |
| mc_nair | smcginnis: thanks for the all the info. Couldn't come up with a good answer for the "why" but the prioritizing makes sense. Was also wondering about why we couldn't just move the "_LW" logic into the LOG.warning method itself, since I think all warning messages get translated? But I guess that might just complicate things since we'd still want LE for the cases where we save the translated message to a variable so we can log and throw it. | 17:11 |
| hemna | mc_nair, you don't want to do that | 17:12 |
| *** vignesh has quit IRC | 17:12 | |
| hemna | not every call to the LOG.* should have a marker wrapped | 17:13 |
| hemna | http://docs.openstack.org/developer/oslo.i18n/guidelines.html#choosing-a-marker-function | 17:13 |
| *** cbader has joined #openstack-cinder | 17:13 | |
| hemna | which is all part of the madness of the markers | 17:15 |
| Swanson | jgriffith, get_replication_updates question. Is there a way to know what volumes I need to return information about without doing a db call to get all volumes associated with my backend? | 17:15 |
| hemna | Swanson, I think the get_replication_updates is a v1 api ' | 17:15 |
| jgriffith | mc_nair: that's the same path I went down when all that marker stuff started :( | 17:15 |
| *** kragniz_ is now known as kragniz | 17:16 | |
| Swanson | hemna, Ah. I thought everything on main was v2. | 17:16 |
| hemna | no, :( | 17:16 |
| hemna | there is some v1 stuff still around | 17:17 |
| hemna | which will get removed once IBM updates their drivers to v2 | 17:17 |
| hemna | jungleboyj, | 17:17 |
| jungleboyj | hemna: Yeah, I don't like how not everything has a marker. | 17:17 |
| jungleboyj | hemna: Yes, in fact baumann1 and jgregor were talking about that this morning in our weekly meeting. | 17:18 |
| mc_nair | jgriffith: yea, I can see how it would lead to it's own complications, as you'd still need the _LE marker because there will be cases that we want to hold onto the message in a var so we can log and then also raise an error. Though I guess that could be it's own helper function | 17:18 |
| hemna | heh, don't get me started on the markers. | 17:18 |
| jungleboyj | hemna: But you love the subject so! | 17:18 |
| mc_nair | haha, I seem to be poking a wound | 17:19 |
| mc_nair | hemna: just trying to understand, so not every call to LOG.* should have a marker, but shouldn't every call to LOG.warning take the _LW marker? | 17:20 |
| mc_nair | I just read that link but I definitely may still be missing something | 17:20 |
| hemna | no | 17:20 |
| hemna | the guidelines provide examples when you should/shouldn't use them | 17:20 |
| hemna | if you have a var that is passed in, instead of a string, don't use the marker | 17:21 |
| hemna | this is part of the insanity of the markers and logging. | 17:21 |
| jungleboyj | mc_nair: Not so much a wound ... we did a lot of work to get this all in place a couple of releases ago. The way it was decided to implement it isn't anyone favorite. | 17:22 |
| Swanson | jgriffith, any outstanding patches for replication I should be aware of? | 17:22 |
| hemna | mc_nair, it's just one of those things that since other OS projects jumped off the cliff with this, we had to as well. | 17:22 |
| *** kevincarr1991 has quit IRC | 17:25 | |
| *** kevincarr1991 has joined #openstack-cinder | 17:26 | |
| mc_nair | alright, still don't 100% get it, but more than I did before. Will try to understand the nuances so I can get bitter about it too :) | 17:26 |
| mc_nair | jgriffith, hemna - thanks for the background | 17:26 |
| hemna | mc_nair, welcome to the club. almost no one 'gets it' wrt the markers, which is why they are so awesome. | 17:26 |
| jgriffith | Swanson: not really :) | 17:29 |
| *** Yogi1 has quit IRC | 17:29 | |
| jgriffith | hemna: do you have a link to our youtube channel with the sessions? | 17:30 |
| hemna | yah | 17:30 |
| diablo_rojo | jgriffith: https://www.youtube.com/channel/UCJ8Koy4gsISMy0qW3CWZmaQ | 17:30 |
| jgriffith | Swanson: the only thing I would like to propose is that things like failover are removed :) | 17:30 |
| hemna | yah that :) | 17:30 |
| jgriffith | diablo_rojo: thanks! | 17:30 |
| *** jdurgin has joined #openstack-cinder | 17:30 | |
| diablo_rojo | jgriffith: hemna: I got you :) | 17:30 |
| Swanson | jgriffith, I can only pledge a +1 to that but that +1 I will pledge. | 17:31 |
| *** dobson has quit IRC | 17:31 | |
| *** hemna is now known as hemnafk | 17:31 | |
| jgriffith | Swanson: I think the only one we have to convince is DuncanT :) | 17:32 |
| *** hemnafk is now known as hemna | 17:33 | |
| Swanson | jgriffith, He isn't in my hemisphere, I don't think. So I can't send anyone to persuade him. | 17:34 |
| jgriffith | sighh... I really wish I would've been at the meetup-1, there's some misunderstandings about replication | 17:34 |
| kevincarr1991 | is there a way to uninstall cinder? | 17:34 |
| jgriffith | and attempts to boil the ocean :( | 17:34 |
| jgriffith | kevincarr1991: why would you want to do such a thing :) | 17:34 |
| jgriffith | kevincarr1991: yes | 17:35 |
| jgriffith | apt-get remove, or yum | 17:35 |
| kevincarr1991 | ha i have made a few errors | 17:35 |
| jgriffith | whatever package manager you use | 17:35 |
| jgriffith | kevincarr1991: tricky part is cleaning up the database and message queue | 17:35 |
| *** akerr_ has quit IRC | 17:35 | |
| *** jistr has quit IRC | 17:36 | |
| *** salv-orlando has quit IRC | 17:39 | |
| jungleboyj | smcginnis: Kilo is security fix only now. Right? | 17:39 |
| *** lcurtis has joined #openstack-cinder | 17:43 | |
| *** Yogi1 has joined #openstack-cinder | 17:45 | |
| *** jordanP has quit IRC | 17:46 | |
| openstackgerrit | Walter A. Boring IV (hemna) proposed openstack/os-brick: Add new Connector APIs for path validation https://review.openstack.org/199764 | 17:52 |
| *** bswartz has quit IRC | 17:55 | |
| *** bswartz has joined #openstack-cinder | 17:56 | |
| *** p0rtal has joined #openstack-cinder | 17:56 | |
| *** willsama has quit IRC | 17:56 | |
| *** willsama has joined #openstack-cinder | 17:57 | |
| *** jgregor has quit IRC | 17:58 | |
| *** jwcroppe has quit IRC | 18:01 | |
| *** jwcroppe has joined #openstack-cinder | 18:01 | |
| *** kevincar_ has joined #openstack-cinder | 18:02 | |
| hemna | patrickeast, ping | 18:04 |
| *** kevincarr1991 has quit IRC | 18:05 | |
| *** jwcroppe has quit IRC | 18:05 | |
| *** salv-orlando has joined #openstack-cinder | 18:08 | |
| *** sghanekar_ has joined #openstack-cinder | 18:09 | |
| *** akerr has joined #openstack-cinder | 18:13 | |
| xyang1 | tbarron: ping | 18:14 |
| tbarron | xyang1: pong | 18:14 |
| xyang1 | tbarron: question for you, are you using SnapMirror for replication | 18:15 |
| tbarron | xyang1: well, I'd like to :-) | 18:15 |
| tbarron | xyang1: why? | 18:16 |
| xyang1 | tbarron: I am trying to understand your use cases:) | 18:16 |
| tbarron | xyang1: makes sense. | 18:16 |
| xyang1 | tbarron: can snapmirror make sure data consistency | 18:17 |
| tbarron | xyang1: yes, the destination has a crash-consistent snapshot of the source filesystem. | 18:18 |
| tbarron | xyang1: in cinder's case, that means of all the volumes in that pool. | 18:19 |
| tbarron | and of all snapshots of those volumes. | 18:19 |
| xyang1 | tbarron: so it is probably ok to use consistencygroup? Just make sure it is in the same pool | 18:19 |
| tbarron | xyang1: consistency group is actually more general, can span pools. | 18:20 |
| tbarron | xyang1: and it can be a proper subset of pool. | 18:21 |
| xyang1 | tbarron: we can make it to work for a pool or across | 18:21 |
| tbarron | xyang1: and for some of our platforms we don't have this snap mirror model. | 18:21 |
| xyang1 | tbarron: not everyone can support cg span pools | 18:21 |
| tbarron | so replication groups are a distinct concept from consistency groups. | 18:21 |
| xyang1 | tbarron: so you have two models, in one pool or span pools? | 18:22 |
| *** hodos has joined #openstack-cinder | 18:22 | |
| *** harlowja has joined #openstack-cinder | 18:22 | |
| tbarron | our CG approach is still getting designed :-) as we have multiple platforms and want to implement in a way that makes sense across them if possible. | 18:23 |
| *** timcl1 has joined #openstack-cinder | 18:23 | |
| tbarron | there are several things we could do to implement CGs. Our pools are already trivially CGs, just not exposed as such. | 18:24 |
| tbarron | But adding/removing individual volumes from them may not make much sense, etc. | 18:24 |
| hemna | tbarron, so isn't that your driver's issue then? | 18:24 |
| *** timcl has quit IRC | 18:25 | |
| tbarron | hemna: it's not just our driver's issue that CGs and Replication Groups are distinct concepts. | 18:25 |
| tbarron | I was just talking about our driver to answer xyang's questions about same. | 18:26 |
| xyang1 | tbarron: there is cg for replication and cg for snapshot | 18:26 |
| hemna | so I guess this is one of the issues with trying to create a general purpose SDS API in Cinder that has to work with many different backends | 18:26 |
| tbarron | hemna: xyang1: my insistence that replication groups are distinct from CGs conceptually isn't really | 18:26 |
| tbarron | driven by just our drivers. | 18:26 |
| xyang1 | tbarron: for lots of drivers they are two groups | 18:27 |
| hemna | we all have to figure out how to map our backend capabilities with the Cinder API concepts | 18:27 |
| tbarron | glusterfs, zfs, probably others are difft. | 18:27 |
| xyang1 | tbarron: for pure, i think they are the same | 18:27 |
| tbarron | xyang1: patrickeast: when I look at pure implementation, there are these things called Pgroups. | 18:28 |
| xyang1 | tbarron: however consistencygroups db in cinder is not soecifically for snapshot or replication | 18:28 |
| xyang1 | It is general | 18:28 |
| tbarron | Pgroups can get replication type attributes, or consistency group type attributes, or both. | 18:28 |
| xyang1 | tbarron: yes pure can be both | 18:28 |
| xyang1 | tbarron: it is most straight forward fir them I think | 18:29 |
| hemna | we have replication groups as well that's not the same as a cg | 18:29 |
| tbarron | xyang1: and I have no objection whatsoever to any vendor having replication groups and consistency groups co-incide. | 18:29 |
| xyang1 | tbarron: for others, we can still use cg table because it is not just for snapshot | 18:29 |
| hemna | but it doesn't preclude a CG from replicating as well | 18:29 |
| tbarron | hemna: ++ | 18:29 |
| tbarron | I have no objection to replicating CGs. Hope that's clear. | 18:30 |
| xyang1 | tbarron: hemna cgsnapshot table is for cg snapshot | 18:30 |
| hemna | yup | 18:30 |
| hemna | we shouldn't mix the 2 | 18:31 |
| tbarron | Just don't want the only kind of replication groups to be CGs. | 18:31 |
| hemna | if we want replication groups in cinder, then that's a separate feature IMHO | 18:31 |
| tbarron | hemna: ++ | 18:31 |
| hemna | that being said, we need to get replication for volumes done/working and implemented by many drivers first. | 18:31 |
| tbarron | hemna: xyang1: And I think it's a pretty simple extension to jgriffith's replication 2.0 to get there. | 18:32 |
| hemna | I don't think we should be forging ahead as quickly with these without more drivers implementing them. | 18:32 |
| xyang1 | hemna: tbarron I am just trying to understand netapp's case because they can't do volume replication | 18:33 |
| *** jwcroppe has joined #openstack-cinder | 18:33 | |
| tbarron | hemna: my concern ATM isn't to rush in an implementation of replication groups but rather to map out the problem space | 18:33 |
| hemna | coolio | 18:33 |
| tbarron | xyang1: we replicate the whole pool with all its volumes. | 18:34 |
| hemna | tbarron, you can't replicate a single volume ? | 18:34 |
| tbarron | xyang1: others might have other sorts of groups. | 18:34 |
| tbarron | hemna: xyang1: so if we failover, we failover the whole pool. | 18:35 |
| xyang1 | tbarron: that is fine, data in your pool is still consistent? | 18:36 |
| tbarron | hemna: xyang1: the RepV2 spec says "that's cool" - there's a paragraph that shouts out to us and others with this kind of setup. | 18:36 |
| tbarron | xyang1: hemna: but the methods in that spec currrently all target individual volumes. | 18:36 |
| tbarron | xyang1: yes, the data is consistent when replicated. | 18:36 |
| hemna | looks like we only have 6 drivers that do CG ? | 18:37 |
| hemna | maybe my search isn't finding all of them | 18:37 |
| hemna | ack-grep ConsistencyGroupVD | 18:37 |
| hemna | ok 8 | 18:37 |
| xyang1 | hemna: 6 vendors | 18:37 |
| hemna | I don't see EMC in that list | 18:37 |
| hemna | hrmm | 18:38 |
| xyang1 | hemna: we didn't do VD | 18:38 |
| hemna | ugh | 18:38 |
| hemna | really ? | 18:38 |
| tbarron | xyang1: hemna: part of the problem with ConsistencyGroupVD is it isn't 100% clear which methods are to be implemented for CGs. | 18:38 |
| tbarron | add/remove vol from CG for example. | 18:39 |
| hemna | the VD has 4 methods | 18:39 |
| xyang1 | tbarron: that is the problem with VD | 18:39 |
| * tbarron is setting Cory up :-) | 18:39 | |
| hemna | create/delete_cg, create/delete_cgsnap | 18:40 |
| xyang1 | hemna: unless every one implements it, we cannot add to VD | 18:40 |
| hemna | *sigh* | 18:40 |
| hemna | then why are we even bothering with the VD mess then? | 18:40 |
| xyang1 | hemna: new methods are update, create from src | 18:40 |
| hemna | I don't get this at all | 18:40 |
| xyang1 | hemna: we could add them and just pass | 18:41 |
| hemna | I dunno, this whole VD thing is broken IMHO | 18:41 |
| xyang1 | hemna: that is the problem. It is too perfect | 18:41 |
| hemna | if they aren't required, then don't make them abc | 18:41 |
| xyang1 | hemna: once you add a new method, it breaks | 18:41 |
| tbarron | be back in 10-15 | 18:42 |
| xyang1 | hemna: you can't make anything required until everyone implements it | 18:42 |
| hemna | I'm with jgriffith on this one. I really don't like the whole VD mess that we have in driver.py | 18:42 |
| hemna | it's not helping clean anything up and it greatly complicates when is supposed to be done by a driver | 18:43 |
| Swanson | "replication groups"? | 18:43 |
| *** stevemar_ has joined #openstack-cinder | 18:46 | |
| *** setmason has joined #openstack-cinder | 18:47 | |
| * tbarron is back | 18:48 | |
| *** akerr has quit IRC | 18:49 | |
| *** stevemar_ has quit IRC | 18:50 | |
| tbarron | Swanson: the idea is that often it makes sense to replicate a set of volumes to a common destination, and failover the same. | 18:50 |
| *** rady has joined #openstack-cinder | 18:51 | |
| smcginnis | jungleboyj: Yep, security only now. | 18:52 |
| smcginnis | jgriffith: +2 for removing failover | 18:52 |
| *** kevincar_ has quit IRC | 18:52 | |
| *** kevincarr1991 has joined #openstack-cinder | 18:53 | |
| Swanson | tbarron, failover is a nightmare for replication. Anyways, my consistency groups are only replicated in that the volumes in the groups are replicated. The backend has no knowledge of the the cg on the front end. | 18:54 |
| smcginnis | Swanson, tbarron: My problem is - we're not actually failing anything over. | 18:55 |
| smcginnis | The admin still needs to redo things on the destination. It would be easier and safer in a disaster scenario to just have them manage existing and set things up again. | 18:56 |
| smcginnis | At least until we can make failover do something useful other than potential F all their data. | 18:56 |
| tbarron | smcginnis: Swanson: I can *do* a failover, i.e. change the destination to be active, but that happens for the whole group of volumes (in my case, the whole storage pool). | 18:57 |
| xyang1 | Swanson: are you testing failover CG already? | 18:57 |
| smcginnis | tbarron: Sounds even less useful then. :) | 18:57 |
| tbarron | smcginnis: well, in true DR I think it's what one wants to do :) AZ1 is on fire, so I want to move everything to AZ2. | 18:58 |
| Swanson | xyang1, nope. I haven't worried about it yet. Just trying to make sure I'm not missing something. | 18:59 |
| smcginnis | tbarron: At least for us, there's not much to do. If the source is gone, just start using the secondary copy. I suppose that is different for other arrays. | 18:59 |
| tbarron | smcginnis: Swanson: but I have no problem with pulling failover method out for now since there doesn't appear to be | 18:59 |
| *** crose has joined #openstack-cinder | 18:59 | |
| tbarron | a useful common denominator. | 18:59 |
| smcginnis | tbarron: +1 | 18:59 |
| Swanson | tbarron, smcginnis: We can do that, too, but without Nova knowing that we moved the volumes.... | 18:59 |
| openstackgerrit | Michal Dulko proposed openstack/cinder-specs: Fix syntax error in Mitaka specs toc https://review.openstack.org/240967 | 18:59 |
| openstackgerrit | Michal Dulko proposed openstack/cinder-specs: Fix syntax error in Mitaka specs toc https://review.openstack.org/240967 | 19:00 |
| tbarron | Swanson: the vms are going to crash. When they re-attach they should get what was the secondary now though. | 19:00 |
| xyang1 | What's point of doing replication if you can never failover when needed?:) | 19:00 |
| smcginnis | It's a manual recovery process anyway. | 19:00 |
| Swanson | tbarron, smcginnis, jgriffith didn't much care for the failover method and didn't have it in the original spec. OTOH if it is going to expand to something more useful... | 19:01 |
| smcginnis | I would rather crawl first, then walk. What it means to "failover" seems to require more discussion. | 19:02 |
| tbarron | smcginnis: ++ | 19:02 |
| *** akerr has joined #openstack-cinder | 19:02 | |
| *** akerr has quit IRC | 19:03 | |
| tbarron | maybe right now we should just be able to say 'replication = True' as a capability if we have it, but not mandate that we all do it the same way. | 19:03 |
| tbarron | I think that was the intent of jgriffith's spec, it just evolved through reviews to be more prescriptive. | 19:03 |
| xyang1 | I don't agree failover should be removed, but may be I am the minority here | 19:03 |
| Swanson | xyang1, Well all I'm doing is breaking replication and renaming the volume on the new backend so it can be picked up. | 19:04 |
| *** merooney has quit IRC | 19:04 | |
| smcginnis | Assuming the source side is there for the replication to be broken, right? Or you unmap? I forget now. | 19:04 |
| hemna | initiating the failover seems ok to me. the revert after the failover is what's problematic | 19:04 |
| *** akerr has joined #openstack-cinder | 19:04 | |
| Swanson | At that point you have a volume that required replication on a backend without replication. You still need to do a lot of manual work to get things back where they were. | 19:05 |
| Swanson | smcginnis, if the source side is there I delete the replication. If the source side isn't I go looking for a volume. If I should find one (one!) I unmap it from the source and rename. | 19:06 |
| Swanson | (Actually I have a todo on the rename.) | 19:06 |
| *** jgregor has joined #openstack-cinder | 19:06 | |
| *** zul_ has joined #openstack-cinder | 19:08 | |
| *** zul has quit IRC | 19:08 | |
| *** akerr has quit IRC | 19:09 | |
| Swanson | What I like about failover is that it is still the same volume in the DB. | 19:10 |
| hemna | Swanson, +1 | 19:10 |
| hemna | and cinder still has access to it | 19:11 |
| tbarron | assuming that when AZ-1 got fried c-vol is still running :-) | 19:12 |
| tbarron | hemna: I agree though. | 19:12 |
| jungleboyj | smcginnis: Yeah, I found the website to confirm it. For some reason I couldn't make it sound right in my head. | 19:12 |
| Swanson | hemna, Yeah, but if you just used manage_existing you get the volume up and can do whatever import magic you need to do. Maybe to AZ1's replacement. | 19:13 |
| jungleboyj | Though, that seems the pattern today. | 19:13 |
| Swanson | xyang1, do you have your cg replication spec url handy? | 19:13 |
| smcginnis | jungleboyj: Inhaling too much helium from all those balloons? :P | 19:13 |
| hemna | :) | 19:13 |
| xyang1 | Swanson: https://review.openstack.org/#/c/229722/ | 19:13 |
| smcginnis | Swanson: And then the volume is actually on the backend it should be on. | 19:13 |
| jungleboyj | No, helium. I started popping them and got jgregor s spit all over. Eeeew. | 19:14 |
| smcginnis | jungleboyj: That's... disturbing. | 19:15 |
| *** timcl has joined #openstack-cinder | 19:16 | |
| jungleboyj | smcginnis: :-) Welcome to my world. | 19:16 |
| *** timcl1 has quit IRC | 19:16 | |
| Swanson | smcginnis, yeah. Of course the thing is people can still import a replication. Actually can they? If I go to do a manage_existing on a repl dest is that an error? | 19:17 |
| Swanson | jgriffith, ^^ | 19:17 |
| jungleboyj | smcginnis: Go on much of a bug cleaning spree? | 19:19 |
| *** baumann has joined #openstack-cinder | 19:20 | |
| smcginnis | jungleboyj: I took a look at a few. ;) | 19:20 |
| smcginnis | Hopefully I didn't blow up anyones inbox. | 19:20 |
| * jungleboyj is crying at the state of my inbox. | 19:21 | |
| *** baumann1 has quit IRC | 19:21 | |
| patrickeast | hemna: pong | 19:23 |
| patrickeast | tbarron: xyang1: just reading the scrollback, but yea for pure we use protection groups (pgroups) which are always consistent for snapshots and stuff (so like 1:1 mapping with a CG), and they are the thing we can replicate | 19:25 |
| *** clayg has left #openstack-cinder | 19:25 | |
| patrickeast | so our replication strategy for cinder is to create a pgroup that is replicated and add/remove volumes from that group when enabling/disabling replication for the volume | 19:25 |
| openstackgerrit | Nate Potter proposed openstack/cinder: Remove 'refresh' parameter from driver get_stats https://review.openstack.org/233328 | 19:27 |
| tbarron | patrickeast: thanks, that fits with what I see in your code. | 19:27 |
| jgriffith | Swanson: as I pointed out in Tokyo, people will make this as complicated as they want | 19:29 |
| jgriffith | Swanson: IMO this should be much more restrictive than what a lot of people are proposing lately | 19:29 |
| xyang1 | patrickeast: thanks, I think it maps well to your pgroup | 19:29 |
| smcginnis | jgriffith: Agreed | 19:30 |
| *** julim has joined #openstack-cinder | 19:32 | |
| *** merooney has joined #openstack-cinder | 19:33 | |
| *** Yogi1 has quit IRC | 19:35 | |
| j_king | jgriffith: are you advocating for simplicity? ;) | 19:40 |
| *** sgotliv has joined #openstack-cinder | 19:41 | |
| openstackgerrit | Tom Barron proposed openstack/cinder-specs: Scaling backup service blueprint spec https://review.openstack.org/240978 | 19:45 |
| *** salv-orl_ has joined #openstack-cinder | 19:46 | |
| jgriffith | j_king: yes please | 19:48 |
| j_king | jgriffith: heresy! | 19:49 |
| *** salv-orlando has quit IRC | 19:49 | |
| *** DericHorn-HP has joined #openstack-cinder | 19:49 | |
| *** salv-orl_ has quit IRC | 19:51 | |
| jgriffith | j_king: indeed, these days it could be seen that way | 19:51 |
| jgriffith | so it's funny, I just saw this article... from Bill Gates of all people | 19:52 |
| jgriffith | but he listed 5 keys of success: http://paste.openstack.org/show/477814/ | 19:52 |
| j_king | jgriffith: i'm with ya. i was disappointed at a few sessions where new abstractions were the default answer to most technical challenges. | 19:52 |
| jgriffith | I find number 1 and number 5 the most compelling | 19:52 |
| jgriffith | j_king: yeah, 10 pounds of crap in a 5 pound bag anyone :) | 19:53 |
| j_king | not that abstractions can't make life more simple... but yeah. | 19:53 |
| j_king | the *right* abstractions tend to have that effect. | 19:53 |
| *** Yogi1 has joined #openstack-cinder | 19:53 | |
| hemna | but if you reverse 5 and 1 you get: "The ability to focus on a goal...no." | 19:53 |
| jgriffith | hemna: ? | 19:53 |
| hemna | </sarcasm> | 19:53 |
| *** jwcroppe has quit IRC | 19:54 | |
| jgriffith | item 1 being most important IMO | 19:54 |
| j_king | indeed. | 19:54 |
| *** jwcroppe has joined #openstack-cinder | 19:54 | |
| *** Yogi11 has joined #openstack-cinder | 19:59 | |
| *** Yogi1 has quit IRC | 20:01 | |
| *** kevincar_ has joined #openstack-cinder | 20:02 | |
| openstackgerrit | Tom Barron proposed openstack/cinder-specs: Scaling backup service blueprint spec https://review.openstack.org/240978 | 20:02 |
| *** DericHorn-HP has quit IRC | 20:02 | |
| *** kevincarr1991 has quit IRC | 20:05 | |
| *** merooney has quit IRC | 20:11 | |
| kevincar_ | Has anyone gotten this error for cinder? WARNING keystonemiddleware.auth_token [-] Authorization failed for token | 20:15 |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/223803 | 20:17 |
| openstackgerrit | Scott DAngelo proposed openstack/cinder-specs: cinder-api-microversions https://review.openstack.org/223803 | 20:26 |
| *** dims has quit IRC | 20:31 | |
| *** crose has quit IRC | 20:33 | |
| *** sghanekar_ has quit IRC | 20:41 | |
| *** stevemar_ has joined #openstack-cinder | 20:46 | |
| *** sghanekar_ has joined #openstack-cinder | 20:47 | |
| *** kevincar_ has quit IRC | 20:49 | |
| smcginnis | scottda: I just needed to officially accept it for M, but I set a priority too to make sure it gets visibility. | 20:50 |
| *** stevemar_ has quit IRC | 20:52 | |
| *** baumann has quit IRC | 20:55 | |
| *** merooney has joined #openstack-cinder | 20:56 | |
| *** merooney has quit IRC | 20:57 | |
| *** salv-orlando has joined #openstack-cinder | 20:59 | |
| *** porrua has quit IRC | 21:06 | |
| *** salv-orlando has quit IRC | 21:08 | |
| scottda | smcginnis: Thanks | 21:09 |
| smcginnis | scottda: No problem. Looks like LP is kind of screwy on that depending if I target then accept, or accept then target. Not the best tool, IMO. :) | 21:10 |
| scottda | I'm glad I'm not the only one who gets confused. | 21:10 |
| *** jamielennox is now known as jamielennox|away | 21:16 | |
| *** dustins has quit IRC | 21:20 | |
| *** timcl has quit IRC | 21:20 | |
| *** kevincarr1991 has joined #openstack-cinder | 21:21 | |
| rlucio | hey guys, @vmem we want to drop driver support for our 6000 series products, can I just submit a review to remove them, or do I need to follow some sort of deprecation route? | 21:22 |
| *** salv-orlando has joined #openstack-cinder | 21:23 | |
| rlucio | by "them" i mean the related driver files | 21:23 |
| *** e0ne has joined #openstack-cinder | 21:23 | |
| hemna | rlucio, I think that's up to you. I presume you are the maintainer of the drivers ? | 21:23 |
| smcginnis | rlucio: You should probably deprecate it for a release. | 21:23 |
| rlucio | hemna: yea i am | 21:23 |
| hemna | rlucio, we have deprecated one of our drivers for a few releases | 21:23 |
| smcginnis | rlucio: Unless you're sure you don't have any customers using it anymore. | 21:23 |
| hemna | smcginnis, the downside to deprecation is that CI has to continue to be running | 21:24 |
| rlucio | smcginnis: for deprecation, is there anything to do in the code? I already have it marked in the docs for deprecation | 21:24 |
| smcginnis | hemna: True | 21:24 |
| hemna | rlucio, might want to output a LOG.warning at do_setup() time | 21:24 |
| rlucio | hemna: exactly, from my seat, less CI maintenance, the better! | 21:24 |
| *** Lee1092 has quit IRC | 21:24 | |
| smcginnis | rlucio: Well, they're your customers, so if you don't want to give a migration window then feel free to just submit a patch to remove them. | 21:25 |
| *** Lindis has joined #openstack-cinder | 21:25 | |
| smcginnis | rlucio: Or stop your CI and I'll do it eventually. ;) | 21:25 |
| smcginnis | kidding | 21:25 |
| *** Lindis has left #openstack-cinder | 21:25 | |
| hemna | lol | 21:25 |
| *** kevincarr1991 has quit IRC | 21:25 | |
| rlucio | smcginnis: well.. i'll double check before removal but before i get there i just wanted to be sure there wasnt a formal process | 21:26 |
| *** Lindis has joined #openstack-cinder | 21:26 | |
| rlucio | smcginnis: haha i like that, kinda of a stealth driver removal... i just stop the ci, and the problem takes care of itself :) | 21:27 |
| smcginnis | rlucio: Normally we deprecate things for a cycle before removing them, but since this is your driver I think you have a little leeway in how it's done. | 21:27 |
| smcginnis | ;) | 21:27 |
| rlucio | hemna, smcginnis: ok maybe for starters i'll add the deprecation warning log msgs, and see what the powers-that-be want to do on the files, given the general method for deprecating for a release | 21:28 |
| openstackgerrit | Walter A. Boring IV (hemna) proposed openstack/cinder: Move get_by_id to CinderObject https://review.openstack.org/218012 | 21:29 |
| *** baumann has joined #openstack-cinder | 21:31 | |
| *** kevincarr1991 has joined #openstack-cinder | 21:31 | |
| *** Lindis has quit IRC | 21:32 | |
| *** apoorvad has joined #openstack-cinder | 21:35 | |
| *** bswartz has quit IRC | 21:38 | |
| *** apoorvad has quit IRC | 21:38 | |
| sghanekar_ | For a documentation merge in the default branch of openstack-manual, when are the changes visible on docs.openstack.org? | 21:41 |
| *** dims has joined #openstack-cinder | 21:41 | |
| *** rady has quit IRC | 21:43 | |
| *** gouthamr has quit IRC | 21:45 | |
| *** can8dnSix has joined #openstack-cinder | 21:48 | |
| *** sgotliv has quit IRC | 21:49 | |
| *** garthb_ has joined #openstack-cinder | 21:50 | |
| *** garthb has quit IRC | 21:52 | |
| *** can8dnSix has quit IRC | 21:52 | |
| *** can8dnSix has joined #openstack-cinder | 21:52 | |
| *** dobson has joined #openstack-cinder | 21:52 | |
| *** salv-orlando has quit IRC | 21:55 | |
| *** rady has joined #openstack-cinder | 21:56 | |
| mtanino | sghanekar_: When I posted a patch https://review.openstack.org/#/c/230863/ , the web was updated soon after the patch was merged. | 21:58 |
| hemna | scottda, ping | 22:00 |
| scottda | Good morning | 22:00 |
| scottda | hemna: | 22:00 |
| hemna | so it looks like the only way to get the initiator info saved in the volume attachment | 22:01 |
| *** e0ne has quit IRC | 22:01 | |
| hemna | is to modify nova | 22:01 |
| hemna | and the cinder api | 22:01 |
| hemna | it has to be done a volume_manager.attach_volume time | 22:01 |
| hemna | and that doesn't accept the connection_info | 22:01 |
| scottda | Yeah, I was thinking that there won't be a clean way to detach from the cinder side unless we detach all attachments for a volume...which is NOT clean. | 22:02 |
| hemna | :( | 22:02 |
| *** aix has joined #openstack-cinder | 22:03 | |
| scottda | There might be a case for that API. Do a force_detach_all from the cinder side in the case where nova has already deleted the VM | 22:03 |
| hemna | so I'll have to add a new column to the attachments table | 22:03 |
| hemna | and then do the song and dance with the API, then client, the nova. | 22:03 |
| *** edmondsw has quit IRC | 22:04 | |
| *** salv-orlando has joined #openstack-cinder | 22:04 | |
| hemna | scottda, https://github.com/openstack/nova/blob/master/nova/virt/block_device.py#L312 | 22:04 |
| hemna | will be a microversion bump | 22:05 |
| scottda | Yeah, will definitely take some orchestration. | 22:05 |
| scottda | But, how would the admin know which attachment to delete using only Cinder if the Nova VM is gone? | 22:06 |
| hemna | it can only be found w/ the instance uuid | 22:07 |
| scottda | Does cinder currently do anything with instance uuid? | 22:07 |
| hemna | the attachment_id is tied to the volume_id and either instance_uuid or host_name | 22:07 |
| scottda | I guess the admin/user might have that, or at least the host_name is in the cinder DB. | 22:07 |
| *** diablo_rojo has quit IRC | 22:07 | |
| hemna | scottda, https://github.com/openstack/cinder/blob/master/cinder/volume/manager.py#L855-L862 | 22:08 |
| hemna | that's how it currently finds the attachment | 22:08 |
| hemna | either by instance_uuid or host_name | 22:08 |
| hemna | + volume_id | 22:08 |
| scottda | It would kinda work except in the case where there are 2 of the same volume (mult-attach) to the same host | 22:08 |
| scottda | So when we land Nova multi-attach that case will be problematic | 22:09 |
| scottda | This is if we don't have the instance_id | 22:09 |
| hemna | you can't attach a volume to the same host_name more than once currently | 22:09 |
| hemna | and by host, I mean bare metal or cinder node | 22:09 |
| scottda | right, but with multi-attach we'll be able to. | 22:09 |
| hemna | attachments to nova compute hosts are done via instance_uuid | 22:09 |
| hemna | no | 22:10 |
| scottda | no? ok | 22:10 |
| scottda | Shouldn't you be able to? | 22:10 |
| hemna | that's the current limitation | 22:10 |
| scottda | right, I think ildikov had mentioned that bug as a blocker | 22:10 |
| scottda | for implementing nova side of multi-attach | 22:10 |
| hemna | where? | 22:11 |
| hemna | I haven't seen/heard that | 22:11 |
| scottda | no, I was wrong. Her issue is https://review.openstack.org/#/c/198400/ | 22:11 |
| hemna | yah and that one is wrong | 22:11 |
| hemna | :P | 22:11 |
| *** Yogi11 has quit IRC | 22:11 | |
| hodos | hi guys. I've looked into remotefs driver...it has _fallocate method, but it seems like it's not used anywhere | 22:12 |
| hemna | see my comments on that one | 22:12 |
| hodos | besides glusterfs | 22:12 |
| scottda | I think when we were talking to J pipes a while back during a hangout we'd glossed over the issue of multi-attach to the same host. But we should make sure it works. | 22:12 |
| scottda | For sure will be needed with ironic. | 22:13 |
| hodos | is it there specifically for glusterfs ? | 22:13 |
| hodos | i mean _fallocate | 22:13 |
| hodos | i mean _fallocate() | 22:13 |
| hemna | scottda, you can't attach a volume more than once to the same host. | 22:13 |
| hemna | you can attach it to more than 1 instance | 22:13 |
| hemna | and by attaching to a host, I mean to a cinder node, or a bare metal node | 22:13 |
| scottda | Right, but more than one instance on the same host should work. | 22:13 |
| hemna | yes, that's not the same thing. | 22:13 |
| scottda | you mean a c-vol node. | 22:14 |
| jgriffith | hemna: just to make sure I follow: Multi-attach is an Instance construct only? | 22:14 |
| hemna | if you have 1 compute host with 10 vms on it, you can attach the same volume to all 10 vms. | 22:14 |
| jgriffith | hemna: it's not "really" multiple iSCSI attachements, it's just one iSCSI to one compute node? | 22:14 |
| hemna | jgriffith, it can multi-attach to instances or hosts. | 22:14 |
| jgriffith | hemna: sorry, but you're contradicting yourself | 22:14 |
| hemna | heh | 22:14 |
| *** salv-orlando has quit IRC | 22:14 | |
| hemna | it seems like it, but I'm not | 22:14 |
| *** diablo_rojo has joined #openstack-cinder | 22:15 | |
| hemna | I'm just not explaining it well | 22:15 |
| hemna | there are 2 use cases | 22:15 |
| jgriffith | hemna: a difference without a distinction :) | 22:15 |
| jgriffith | hemna: but I think I'm with you | 22:15 |
| *** salv-orlando has joined #openstack-cinder | 22:15 | |
| hemna | 1) attaching a volume to multiple instances. Those instances can be on the same compute host or multiple compute hosts. | 22:15 |
| *** mdenny has joined #openstack-cinder | 22:15 | |
| hemna | 2) attaching a volume to a host. | 22:15 |
| hemna | and by host, I mean c-vol or bare metal node. | 22:16 |
| jgriffith | hemna: then say "node" | 22:16 |
| jgriffith | :) | 22:16 |
| hemna | ok | 22:16 |
| jgriffith | hemna: but how is 2 relevant to multi-attach discussions? | 22:16 |
| hemna | sure. unfortunately, the db table colume is attached_host | 22:17 |
| hemna | jgriffith, for allowing attachments to BM nodes, via the brick-client | 22:17 |
| hemna | either way | 22:17 |
| jgriffith | hemna: I seem to recall recommending attached_hosts and making that a list... but regardless, still not seeing your point | 22:17 |
| hemna | we have to save the initiator connector info in the attachment row in the db. | 22:18 |
| hemna | my point is we can't do it currently w/o api changes | 22:18 |
| jgriffith | hemna: can't do "what" ? | 22:18 |
| hemna | to allow passing in the connector info at volume manager attach_volume() | 22:18 |
| hemna | as it doesn't have that param currently | 22:18 |
| jgriffith | hemna: why don't you just use the connector from initialize_connection? | 22:20 |
| jgriffith | hemna: still another tagnetial thing | 22:20 |
| scottda | jgriffith: We were hoping for a way to force_detach a volume from only the cinder side, used in the case when the nova instance has been deleted but the cinder volume and connection has not been cleaned up. | 22:20 |
| jgriffith | scottda: understood | 22:20 |
| scottda | But cinder does not keep the connection info... | 22:20 |
| hemna | because initialize_connection is called outside of the work flow of attaching a volume | 22:20 |
| hemna | re: live migration time. | 22:21 |
| jgriffith | hemna: huh? | 22:21 |
| jgriffith | hemna: I beg to differ | 22:21 |
| hemna | nova calls initialize_connection several times in the process of live migration | 22:21 |
| jgriffith | hemna: as there are drivers that require the initiator iqn to do attachments | 22:21 |
| jgriffith | hemna: I'm now completely confused | 22:22 |
| hemna | and nova isn't asking cinder to attach the volume, but to ask it for the response | 22:22 |
| jgriffith | hemna: none of this has anything to do with the question that was asked earlier that I can see regarding what one can and can't multi-attach too | 22:22 |
| hemna | ok, well I guess we are talking about 2 things then | 22:22 |
| hemna | I'm trying to find out how to get the connector info stuffed into the volume attachment table | 22:23 |
| hemna | so we can do force detach | 22:23 |
| hemna | and baremetal attaches w/ brick-client | 22:23 |
| hemna | as we discussed in the meetup in tokyo | 22:23 |
| hemna | and the only way to do it is to update the cinder api to accept connector in attach_volume time | 22:23 |
| jgriffith | hemna: well, now "we", I just asked what the deal was when you said "you can only multi-attach to instances on the same compute node" then said "you can attach to multiple compute nodes" then said "something blah blah about bare-metal" | 22:23 |
| * jgriffith is rather confused | 22:23 | |
| openstackgerrit | Sonia Ghanekar proposed openstack/cinder: Using extra-specs in cloned vols for Nimble driver https://review.openstack.org/237735 | 22:24 |
| jgriffith | hemna: do you want to talk about that or the original question scottda asked? | 22:24 |
| jgriffith | hemna: I'm happy to discuss either :) | 22:24 |
| scottda | I think the main problem is that the clocks says it is aftenoon, but in Tokyo it is 7:24 AM. Perhaps we should all sleep on it. | 22:24 |
| jgriffith | scottda: ok.. but I strongly urge folks to reconsider the question around initiator data | 22:25 |
| smcginnis | hemna, scottda: Remind me again why we need to know the initiator to force a delete? | 22:25 |
| hemna | jgriffith, what do you mean ? | 22:25 |
| scottda | We need the connector. | 22:26 |
| smcginnis | scottda: And why? | 22:26 |
| hemna | smcginnis, becaused in order to correctly call terminate_connection in the driver, you have to have the connector | 22:26 |
| jgriffith | hemna: I mean I don't think this problem is as difficult as it's being made out to be, and scottda we have the connector | 22:26 |
| scottda | And I might have said delete. That would be a mistake. I meant detach | 22:26 |
| hemna | smcginnis, or the driver can't know what to unexport from the array. | 22:26 |
| smcginnis | hemna: So we aren't deleting, just detaching? | 22:26 |
| hemna | smcginnis, correct, this is force detach | 22:27 |
| smcginnis | hemna: OK, thanks. So we need to know where it's exported to in the case of multiple attach? | 22:27 |
| hemna | jgriffith, it's not difficult, just needs an API bump to add the connector to attach_volume | 22:27 |
| smcginnis | So we don't detach from something else that is still there and using it? | 22:27 |
| jgriffith | hemna: and I'm arguing that I don't think that's necessary | 22:27 |
| scottda | Any detach requires a connector. It is passed in by nova, but only if nova has the info. It does not if the nova instance has already been deleted, which sometimes happens | 22:28 |
| hemna | jgriffith, ok, well that's the only way I see it working correctly | 22:28 |
| jgriffith | hemna: https://github.com/openstack/cinder/blob/master/cinder/volume/targets/lio.py#L174 | 22:28 |
| hemna | sure | 22:28 |
| smcginnis | scottda: So if we're forcing the detach, do we care where it was? Why not just remove everything for that volume? | 22:29 |
| hemna | the problem is for detaching when the vm instance is gone, we don't have the connector anywhere. | 22:29 |
| *** mriedem is now known as mriedem_away | 22:29 | |
| hemna | nova doesn't have it, and cinder doesnt' have it | 22:29 |
| *** julim has quit IRC | 22:29 | |
| scottda | jgriffith: That is a problem with multi-attach. YOu might not want to detach from all instances.... | 22:29 |
| hemna | smcginnis, we can't find which export to remove unless we have the initiator information. | 22:29 |
| *** baumann has left #openstack-cinder | 22:29 | |
| jgriffith | hemna: scottda umm, but I think we do | 22:29 |
| scottda | but you could do a force_detach_all . It might be the only way in the case of multi- attach, and would require the changes hemna has proposed (keeping the info) | 22:30 |
| jgriffith | hemna: scottda IIRC patrickeast added something like this a while back | 22:30 |
| *** vgridnev has quit IRC | 22:30 | |
| hemna | scottda, correct, we don't want to remove every attachment for that volume every time. | 22:30 |
| jgriffith | hemna: scottda because he needed initiator info in the db for his access groups to make sure he didn't recreate/duplicate | 22:30 |
| hemna | scottda, +1 | 22:30 |
| smcginnis | Maybe it's a difference in how arrays work. But if I'm told to force detach a volume, I could just remove whatever is configured for that volume without needing to care. | 22:30 |
| smcginnis | Multiattach I could see be an issue though. | 22:30 |
| hemna | smcginnis, and when that volume is attached to more than one thing ? | 22:30 |
| hemna | :) | 22:30 |
| *** jamielennox|away is now known as jamielennox | 22:31 | |
| hemna | so the connector is what the volume is exported to, and to not wipe out all exports for that volume, you need that connector (initiator info) | 22:31 |
| scottda | Yeah, so the question is "do we want a fine-grained force-detach for a single atttachment for multi-attach cases?" and "can we even do this?" | 22:31 |
| patrickeast | jgriffith: this thing https://github.com/openstack/cinder/blob/master/cinder/db/sqlalchemy/models.py#L590 ? | 22:31 |
| hemna | scottda, yes | 22:31 |
| jgriffith | patrickeast: ahh, yeah, that's what I was thinking of | 22:31 |
| jgriffith | patrickeast: IIRC you're already doing what scottda and hemna are proposing no? | 22:32 |
| hemna | patrickeast, in the case of multi-attach you will have different connector info for each attachment | 22:32 |
| jgriffith | patrickeast: scottda hemna I also think that provider data is a descent enough place for this sort of thing | 22:32 |
| hemna | the connector needs to be stored with the specific attachment in the attachments table to be correct. | 22:33 |
| jgriffith | patrickeast: scottda hemna when the volume is attached return a model update that populates the initiators | 22:33 |
| jgriffith | but anyway | 22:33 |
| jgriffith | hemna: why? Why not just store it with the volume? | 22:33 |
| jgriffith | hemna: if you just want it for force detach/detach all, then why do you care? | 22:33 |
| hemna | because the connector will be different for every attachment | 22:34 |
| jgriffith | hemna: as long as you have all of them, why make it harder and introduce yet another source of truth regarding attachment info | 22:34 |
| jgriffith | hemna: yes, I understand that | 22:34 |
| hemna | there is 1 source of truth, it's the volume_attachment table | 22:34 |
| hemna | the connector should go in there for each attachment | 22:34 |
| jgriffith | hemna: since you mention that.. I'm also confused by that still | 22:35 |
| hemna | initialize_connection doesn't have the instance_uuid and attach_volume doesn't have the connector | 22:36 |
| hemna | so, I'm proposing updating attach_volume to add the connector | 22:36 |
| jgriffith | hemna: So are you just proposing that we add initiator to the VolumeAttachment table? | 22:36 |
| hemna | attach_volume already has the code to find the correct attachment. | 22:36 |
| hemna | yup | 22:36 |
| jgriffith | hemna: and I'm saying I don't understand why you would do that, honestly. But really I guess I don't care | 22:37 |
| hemna | because that's the correct attachment for that connector | 22:37 |
| jgriffith | hemna: the adding the connector to attach call | 22:37 |
| jgriffith | hemna: ok | 22:37 |
| jgriffith | hemna: I won't frustrate you any more | 22:37 |
| hemna | hehe | 22:37 |
| jgriffith | hemna: just state that I don't think it's necessary and move on :) | 22:37 |
| scottda | We need to talk it all through anyway, and we'll probably do it again in the spec :) | 22:38 |
| hemna | I'm ok with talking it through | 22:38 |
| hemna | I want to understand your viewpoint | 22:38 |
| hemna | usually these discussions end up in having 'aha!' moments for someone either way | 22:38 |
| hemna | usually me. | 22:38 |
| hemna | so, I'd like to be able to store the data at initialize_connection time | 22:39 |
| jgriffith | hemna: yeah, so why not do that? | 22:39 |
| patrickeast | so, after reading the backscroll and what i remember from the ironic session the proposal is simply that we keep the initiator info in the cinder db whenever we do an attach, right? | 22:39 |
| jgriffith | patrickeast: correct | 22:39 |
| hemna | but 1) initialize_connection is called by nova outside of the workflow for attaching a volume | 22:39 |
| patrickeast | then we can look it up and clean up connections | 22:39 |
| patrickeast | or detach ironic | 22:39 |
| patrickeast | or whatever needs to be done | 22:39 |
| patrickeast | is that it? | 22:39 |
| hemna | and 2) we don't have a way to find the right volume_attachment table entry | 22:39 |
| jgriffith | patrickeast: yes... unless hemna and scottda have a case I'm not thinking of | 22:40 |
| hemna | patrickeast, yes and the only place we have of storing that connector with the right attachment is at attach_volume time currently. | 22:40 |
| jgriffith | hemna: wait... let's back up to item 1 | 22:40 |
| jgriffith | hemna: so you say multiple initialize_connections... but so? | 22:40 |
| scottda | I can only think of the 2 cases 1) for ironic detach and 2) force-detach for pathological case when nova instance has been deleted but detach fails for cinder volume | 22:40 |
| hemna | jgriffith, yes, live migration calls initialize_connection several times outside of attaching a volume. | 22:41 |
| jgriffith | hemna: and? The problem is? | 22:41 |
| jgriffith | hemna: if you already have the data, then don't store it, if it's "new" store it | 22:41 |
| hemna | that can update the connector | 22:41 |
| hemna | and incorrectly | 22:41 |
| *** _cjones_ has joined #openstack-cinder | 22:41 | |
| hemna | we had a case where nova was doing a detach against the wrong connector info | 22:42 |
| jgriffith | hemna: right, but you don't have to update, just append | 22:42 |
| jgriffith | hemna: and IMO what you describe is just a bug, if that's what' going on | 22:42 |
| hemna | so, if we knew what instance_uuid/host was being attached at initialize_connection time, then we could find the volume_attachment entry and update it. | 22:42 |
| hemna | jgriffith, well it's not a bug now, we fixed that in nova in L | 22:43 |
| jgriffith | hemna: sadly it's been a while since I've looked at Nova-LiveMigration to remember why it's calling initialize_connection outside of an attach. | 22:43 |
| hemna | but it made me aware that initialize_connection is being called many places in nova | 22:43 |
| *** edtubill has quit IRC | 22:43 | |
| jgriffith | hemna: have you documented any of this or written up the flow that you're concerned about? | 22:43 |
| scottda | BBIAB | 22:44 |
| hemna | no | 22:45 |
| jgriffith | hemna: because when I look at the code it seems like the calls are in fact for attaching back to new dest instance | 22:45 |
| hemna | I was just thinking outloud in here, to get feedback before I went ahead w/ a spec | 22:45 |
| jgriffith | post_live_migration and init_volume_connection | 22:46 |
| *** lprice has quit IRC | 22:46 | |
| hemna | yah | 22:47 |
| hemna | https://github.com/openstack/nova/blob/master/nova/virt/block_device.py#L312 | 22:47 |
| hemna | I believe that's the only place that nova calls cinder's attach api | 22:47 |
| hemna | to finish the attachment | 22:48 |
| jgriffith | hemna: I suppose the refresh might be your concern? | 22:48 |
| hemna | just need to add connector to that call | 22:48 |
| openstackgerrit | Nate Potter proposed openstack/cinder: Added more options while uploading volume as image https://review.openstack.org/230617 | 22:48 |
| jgriffith | hemna: oh... well, that's the only explicit call to cinders attach | 22:48 |
| hemna | yup | 22:48 |
| jgriffith | hemna: oh.. this goes back to your device swapping targets on people? | 22:48 |
| jgriffith | hemna: I mean, you don't reuse or use the same target for the life of the volume? | 22:49 |
| *** can8dnSix has quit IRC | 22:49 | |
| jgriffith | hemna: which was why you changed the workflow before right? | 22:49 |
| hemna | for 3PAR, we export a new target for every attachment | 22:50 |
| hemna | for lefthand 1 target per host | 22:50 |
| jgriffith | hemna: yeah, that's what I'm talking about | 22:50 |
| hemna | it's just how they work. /me shrugs | 22:50 |
| hemna | so I can understand the confusion I guess | 22:51 |
| jgriffith | hemna: so the problem is that 3par when it gets another initialize_connection it creates a new target and the compute node doesn't have that because it expects the target to be the same right? | 22:51 |
| hemna | that was one of our problems yes. | 22:51 |
| hemna | we fixed our driver to be careful with initialize_connection calls | 22:51 |
| jgriffith | hemna: so rather than monkey around too much, shouldn't we just have Nova explciity detach/reattach during live migration for devices that don't keep targets? | 22:52 |
| jgriffith | hemna: that could be encapsulated in provider info | 22:52 |
| jgriffith | hemna: and for 3par nova could explicitly detach/reattach rather than just trying to use the same target info | 22:53 |
| hemna | honestly, I don't know why nova does what it does during live migration | 22:53 |
| jgriffith | hemna: I'd also be curious to look at Pure and LVM with LIO | 22:53 |
| hemna | I tested LVM with live migration and it didn't have the same problem | 22:54 |
| hemna | same w/ lefthand | 22:54 |
| jgriffith | hemna: because they'll have some funny stuff with initiator/access groups | 22:54 |
| hemna | because the target was the same basically. | 22:54 |
| jgriffith | hemna: right... but use LIO :) | 22:54 |
| hemna | hehe | 22:54 |
| hemna | I haven't messed w/ LIO much | 22:54 |
| hemna | but maybe I should | 22:54 |
| hemna | I was curious how LIO did FC | 22:54 |
| jgriffith | hemna: LIO is going to use the concept of initiator access groups IIRC | 22:54 |
| patrickeast | our initialize connection is safe to call multiple times, we handle the case where we get the same initiator more than once | 22:54 |
| jgriffith | hemna: meh, not even getting to that... just LIO iSCSI | 22:55 |
| hemna | :P | 22:55 |
| jgriffith | patrickeast: yeah, that's what I am wondering, why the fix isn't in the driver | 22:55 |
| hemna | patrickeast, coolio. we had a bug in our driver wrt to that. we had to fix in L. | 22:55 |
| patrickeast | kind of seems like making drivers be able to handle it is the easiest approach | 22:55 |
| *** jgregor has quit IRC | 22:55 | |
| * patrickeast says that knowing nothing about how hard it might be for other vendors :p | 22:56 | |
| hemna | the other side of it, in the live migration case, is that the target info wasn't always the same between src host and dest host. | 22:56 |
| jgriffith | patrickeast: seems like it, the other thing is maybe hemna you should just throw up some code that does what you're talking about? | 22:56 |
| hemna | yah I could do that | 22:56 |
| jgriffith | hemna: since chatting and reading specs is always full of peril :) | 22:56 |
| hemna | :P | 22:56 |
| jgriffith | hemna: maybe if I saw the code I'd just be like "Ohhhh SNAP, now I get what you're saying" | 22:57 |
| hemna | I could make this a dep on scottda's microversion patch | 22:57 |
| patrickeast | what i'm wondering though is what live migration has to do with storing connector info in the cinder db for doing force clean-ups | 22:57 |
| hemna | or not, and just put up a WIP POS patch | 22:57 |
| patrickeast | wouldn't we need it either way, live migration or not? | 22:57 |
| jgriffith | patrickeast: yeah, I'm honestly really confused by all of this now TBH | 22:57 |
| hemna | patrickeast, well only tangentially | 22:57 |
| *** jungleboyj has quit IRC | 22:58 | |
| hemna | because of the calling of initialize_connection multiple times throughout the process. | 22:58 |
| patrickeast | yea i mean, its the gun that was used to do the foot-shooting that requires having a force cleanup | 22:58 |
| hemna | normally folks think of initialize_connection being called in the context of doing an attachment | 22:58 |
| hemna | but live migration doesn't | 22:58 |
| patrickeast | gotcha, so for the db entry we need to know if we update it or not | 22:59 |
| *** gouthamr has joined #openstack-cinder | 22:59 | |
| hemna | and if I add the connector to manager.attach_volume() then it's all good. | 22:59 |
| hemna | as nova only calls that in the process of attaching a volume. | 22:59 |
| hemna | here: https://github.com/openstack/nova/blob/master/nova/virt/block_device.py#L312 | 22:59 |
| *** gouthamr_ has joined #openstack-cinder | 23:00 | |
| jgriffith | hemna: I've seen this movie before :) | 23:00 |
| *** daneyon has quit IRC | 23:01 | |
| hodos | has _update_volume_stats changed for iSCSI??? | 23:02 |
| jgriffith | hodos: go to github and look | 23:02 |
| hodos | but this is odd | 23:03 |
| *** gouthamr has quit IRC | 23:03 | |
| *** diablo_rojo has quit IRC | 23:04 | |
| jgriffith | hemna: https://github.com/openstack/cinder/commit/ecab4005f08781581d3d9239e7145fca2a259b35#diff-724c53b8cd580ae533b4fb301df7c3e9 | 23:04 |
| jgriffith | hemna: so you just want to obsolete initialize_connection ? | 23:04 |
| *** alejandrito has quit IRC | 23:04 | |
| hemna | no | 23:05 |
| hemna | I just want to add the connector to attach_volume | 23:05 |
| jgriffith | hemna: it seems weird that we're going to end up passing the connector in to all of these method calls | 23:05 |
| hemna | the other way to do it | 23:05 |
| jgriffith | hemna: sure.... but you've already added it to create_export, now attach and it's already in initialize_connection | 23:05 |
| hemna | is to add instance_uuid/host to initialize_connection | 23:05 |
| *** chlong has joined #openstack-cinder | 23:06 | |
| jgriffith | hemna: Yeah, I just don't get it... but like I said; I shouldn't waste any more of your cycles on it | 23:06 |
| *** smatzek has quit IRC | 23:06 | |
| hemna | coolsvap, I'll just post up a patch then | 23:06 |
| hemna | bah | 23:07 |
| jgriffith | hemna: I'm sure you've looked at this more than I have at this point | 23:07 |
| hemna | bad xchat | 23:07 |
| jgriffith | hemna: and know what you need | 23:07 |
| jgriffith | hehe | 23:07 |
| hodos | jgriffith: so this is inconsistent now for remotefs and block_device | 23:11 |
| *** jwcroppe has quit IRC | 23:11 | |
| hodos | jgriffith: get_volume_stats i mean | 23:12 |
| hodos | jgriffith: remotefs check for self._stats | 23:12 |
| hodos | jgriffith: block_device - doesn't | 23:12 |
| openstackgerrit | Anthony Lee proposed openstack/cinder: Refactor HP 3PAR drivers to now be HPE https://review.openstack.org/240338 | 23:25 |
| *** ociuhandu has quit IRC | 23:26 | |
| *** bswartz has joined #openstack-cinder | 23:27 | |
| openstackgerrit | Anthony Lee proposed openstack/cinder: Refactor HP LeftHand driver to now be HPE https://review.openstack.org/240343 | 23:30 |
| openstackgerrit | Anthony Lee proposed openstack/cinder: Refactor HP 3PAR drivers to now be HPE https://review.openstack.org/240338 | 23:31 |
| *** r-daneel has joined #openstack-cinder | 23:41 | |
| hodos | jgriffith: https://review.openstack.org/#/c/233328/7 i think this review is breaking it | 23:42 |
| *** martyturner has quit IRC | 23:45 | |
| patrickeast | hodos: whats broken with it? | 23:46 |
| tbarron | hodos: just saw your question re. fallocate in the backlog. It does appear to be used only by glusterfs currently. | 23:46 |
| hodos | patrickeast: nothing, sorry, just need to update my driver | 23:46 |
| tbarron | hodos: but I don't see why nfs, etc. couldn't also use it when creating un-sparsed files. Faster than 'dd'. | 23:47 |
| tbarron | eharney is on vacation this week or I'd ask him right now. | 23:47 |
| hodos | patrickeast: actually it's the first time when CI tests showed that i have something wrong in the driver code | 23:47 |
| tbarron | hodos: but why are you asking? (go ahead and finish the other conversation first though :-) | 23:48 |
| *** sileht has quit IRC | 23:48 | |
| * tbarron goes off to eat dinner. | 23:49 | |
| hodos | tbarron: exactly, it looks like we could improve _create_regular_file in remotefs driver | 23:49 |
| hodos | tbarron: maybe not all OS have it? | 23:50 |
| tbarron | hodos: +1 unless there's something I don't see. | 23:50 |
| jgriffith | hodos: why do you think that patch breaks somehting for you? | 23:50 |
| jgriffith | hodos: that isn't even merged yet, so not sure what you're running in to | 23:50 |
| hodos | jgriffith: i run into this during CI | 23:50 |
| jgriffith | hodos: oh, is your CI testing it? | 23:51 |
| jgriffith | hodos: or I should ask "which" CI is yours :) | 23:51 |
| hodos | hodos: yes, my driver is not passing it, i think it's a bug in my driver | 23:51 |
| jgriffith | hodos: which driver is yours? | 23:51 |
| hodos | jgriffith: nexenta/iscsi | 23:51 |
| jgriffith | I see IBM failed, but that one always fails | 23:51 |
| jgriffith | ahh | 23:51 |
| hodos | jgriffith: nexenta ci | 23:51 |
| hodos | jgriffith: yes there's useless code in my iscsi driver | 23:52 |
| jgriffith | hodos: oh... ouch! | 23:52 |
| jgriffith | hodos: ahh... yeah, you're getting the dreaded "No host found" :( | 23:53 |
| hodos | jgriffith: yes, cause the driver got kicked out it hasn't been updated in the patch | 23:54 |
| jgriffith | ahhhh | 23:54 |
| jgriffith | that's a bummer (and a pain) | 23:55 |
| *** garthb_ has quit IRC | 23:55 | |
| *** kevincar_ has joined #openstack-cinder | 23:56 | |
| *** willsama has quit IRC | 23:56 | |
| hodos | jgriffith: well at least CI proved to do some good | 23:57 |
| jgriffith | hodos: indeed! | 23:57 |
| *** garthb_ has joined #openstack-cinder | 23:59 | |
| *** kevincarr1991 has quit IRC | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
