| *** yangyapeng has quit IRC | 00:16 | |
| *** zhurong has joined #openstack-freezer | 00:48 | |
| *** yangyapeng has joined #openstack-freezer | 01:39 | |
| *** yangyapeng has quit IRC | 01:55 | |
| *** Marcellin__ has quit IRC | 01:59 | |
| *** yangyapeng has joined #openstack-freezer | 02:00 | |
| *** c00281451_ has joined #openstack-freezer | 03:05 | |
| *** zengchen has quit IRC | 03:09 | |
| *** zhurong has quit IRC | 03:40 | |
| *** vAlexa_ has joined #openstack-freezer | 05:35 | |
| *** vAlexa has quit IRC | 05:36 | |
| *** lpetrut has joined #openstack-freezer | 06:15 | |
| *** vAlexa_ has quit IRC | 06:49 | |
| *** lpetrut has quit IRC | 06:56 | |
| *** saggi has quit IRC | 07:16 | |
| *** yuvalb has quit IRC | 07:16 | |
| *** lpetrut has joined #openstack-freezer | 07:16 | |
| *** yuvalb has joined #openstack-freezer | 07:17 | |
| *** saggi has joined #openstack-freezer | 07:17 | |
| *** yuvalb has quit IRC | 07:31 | |
| *** yuvalb has joined #openstack-freezer | 07:32 | |
| *** lpetrut has quit IRC | 07:50 | |
| *** yuval has joined #openstack-freezer | 08:01 | |
| *** openstackgerrit has quit IRC | 08:03 | |
| *** lpetrut has joined #openstack-freezer | 08:06 | |
| *** vAlexa has joined #openstack-freezer | 08:24 | |
| *** lpetrut has quit IRC | 08:56 | |
| *** electrofelix has joined #openstack-freezer | 08:56 | |
| *** vAlexa has quit IRC | 09:19 | |
| *** vAlexa has joined #openstack-freezer | 09:21 | |
| *** yangyapeng has quit IRC | 10:18 | |
| *** yangyapeng has joined #openstack-freezer | 11:44 | |
| *** openstackgerrit has joined #openstack-freezer | 12:08 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/freezer: Updated from global requirements https://review.openstack.org/423663 | 12:08 |
|---|---|---|
| *** catintheroof has joined #openstack-freezer | 12:16 | |
| *** lpetrut has joined #openstack-freezer | 12:37 | |
| *** lpetrut has quit IRC | 13:04 | |
| *** AlexeyAbashkin has joined #openstack-freezer | 13:16 | |
| *** daemontool has joined #openstack-freezer | 14:18 | |
| daemontool | all, https://etherpad.openstack.org/p/freezer_cinder-os-brick please take a look | 14:32 |
| daemontool | it would be a new engine | 14:32 |
| daemontool | like cinder-osbrick | 14:33 |
| daemontool | to backup volume, without passing throught Glance | 14:33 |
| daemontool | and leveraging os_bricks from Cinder | 14:33 |
| daemontool | planning to create a spec today for review, thanks | 14:33 |
| slashme | Thx daemontool :) | 14:34 |
| slashme | Seems interresting | 14:34 |
| slashme | I'll take a look :) | 14:34 |
| daemontool | thanks slashme | 14:34 |
| daemontool | I have couple of thought | 14:35 |
| daemontool | was discussing | 14:35 |
| daemontool | s/discussing/thinking/ | 14:35 |
| daemontool | it is possible to mount the image of a VM? | 14:35 |
| raliev | hey everyone :) | 14:35 |
| slashme | daemontool: Yes, through a loopback | 14:35 |
| daemontool | like VM -> snapshot -> (convert to something)? -> mount it | 14:35 |
| slashme | Hi raliev | 14:36 |
| daemontool | not sure if os-brick can be used for VMs | 14:36 |
| daemontool | but we could use the same approach | 14:36 |
| daemontool | also for instances backup | 14:36 |
| daemontool | eventually | 14:36 |
| daemontool | so with os-brick basically we do not pass throught glance | 14:36 |
| daemontool | that's the difference | 14:36 |
| daemontool | the summary of the workflow would be | 14:36 |
| daemontool | volume snap -> convert snap to volume -> attach it with os-brick -> mount it on the vm where the freezer-agent is runnning | 14:37 |
| daemontool | and then execute a normal fs backup of / | 14:37 |
| slashme | I did not know about os-brick | 14:37 |
| *** dstepanenko has joined #openstack-freezer | 14:37 | |
| daemontool | then umount, detach, remove the new volume | 14:37 |
| daemontool | slashme: it's a library that allows you to mount volumes from the vm | 14:38 |
| daemontool | directly | 14:38 |
| slashme | It provides python utilities to access cinder-columes regardless of the cinder backend then ? | 14:38 |
| daemontool | http://docs.openstack.org/developer/os-brick/api/os_brick/initiator/connector.html | 14:38 |
| daemontool | yes | 14:38 |
| daemontool | exactly | 14:38 |
| slashme | That's pretty cool | 14:38 |
| daemontool | but, you do not have access | 14:39 |
| daemontool | to the volume file as a whole | 14:39 |
| daemontool | but you can mount it | 14:39 |
| daemontool | direclty | 14:39 |
| daemontool | and see the content | 14:39 |
| daemontool | se we save the Glance step | 14:39 |
| daemontool | for volumes backup | 14:39 |
| daemontool | and was wondering if the same approch (that of course need to be evaluated by the Team) | 14:39 |
| daemontool | can work for instances too | 14:40 |
| daemontool | anyway all take a look | 14:40 |
| slashme | daemontool, if the ephemeral uses volumes (for example this can be configured with ceph), then, I guess yes. But not in most case | 14:40 |
| daemontool | ok | 14:41 |
| daemontool | the other point I'm wondering is | 14:42 |
| daemontool | what volume clone does underneath | 14:42 |
| daemontool | what happen if the volume is being written while a clone is happening | 14:42 |
| daemontool | because in that spec | 14:42 |
| daemontool | the proposal would be to use | 14:43 |
| slashme | I guess it depends on the backend | 14:43 |
| *** lpetrut has joined #openstack-freezer | 14:43 | |
| daemontool | snap -> convert to vol back (so the volume can be attached with os-brick) | 14:43 |
| daemontool | but snap -> convert back to vol | 14:43 |
| daemontool | may take time | 14:43 |
| daemontool | so was thinking about clone | 14:43 |
| daemontool | ok | 14:43 |
| raliev | daemontool, did you read this article? http://docs.openstack.org/admin-guide/blockstorage-volume-backups.html | 14:54 |
| daemontool | raliev: yes | 14:58 |
| daemontool | thing about that is that cinder-backup needs to be available | 14:58 |
| daemontool | and | 14:58 |
| daemontool | only swift can be used as stoarge | 14:58 |
| daemontool | also all the operations there on volumes, like compress and encrypt | 14:59 |
| daemontool | happens in the storage node | 14:59 |
| daemontool | while our approach with the agent is distributed, it probably scales better | 14:59 |
| daemontool | the other thing is | 14:59 |
| daemontool | that the volumes encryption keys | 14:59 |
| daemontool | are managed by the storage node | 14:59 |
| daemontool | while we give to each tenant his own key | 15:00 |
| daemontool | with the storage brick approach, we backup data within the volume, without having the agent installed in the VM | 15:01 |
| daemontool | so that can be restore anywhere | 15:01 |
| daemontool | on another volume, on a VM non-ephemeral | 15:01 |
| daemontool | on a bare metal | 15:01 |
| daemontool | probably more flexible | 15:02 |
| daemontool | I need to write the restore part thought... | 15:02 |
| daemontool | (in the spec) | 15:02 |
| raliev | so, the most important problem for now - step about creating a snap and converting back to a volume | 15:03 |
| raliev | am I right? | 15:03 |
| daemontool | well that's 2 API calls | 15:05 |
| daemontool | issues is amount of time | 15:05 |
| daemontool | to make that happen | 15:05 |
| daemontool | related to the Volume size | 15:05 |
| daemontool | that's why I was asking before is volume clone | 15:05 |
| daemontool | can be of any help to speed up the process (but consistency can be an issue there is snapshot is not taken) | 15:06 |
| *** yangyapeng has quit IRC | 15:12 | |
| daemontool | raliev: make sense a bit what I'm saying? | 15:31 |
| daemontool | probably some points are weak | 15:31 |
| daemontool | but I think distributed approach is probably better | 15:32 |
| raliev | daemontool, yep, but we'll have a whole picture after describing restore process | 15:32 |
| daemontool | good statement :) | 15:33 |
| raliev | I'll try to ask Ivan Kolodyazhny from cinder team for review this spec, mb he has ideas how to improve that :) | 15:36 |
| daemontool | ok thanks, let's wait couple of dys | 15:38 |
| daemontool | so I can add more contant | 15:38 |
| daemontool | re: restore | 15:38 |
| daemontool | definetely yes, thanks | 15:38 |
| raliev | so, our final goal is to run freezer-agent on root dir of attached volume and create backup file with it's content | 15:39 |
| daemontool | yes | 15:40 |
| daemontool | otherwise we have to implement a driver for each storage backend | 15:41 |
| daemontool | to act directly on volume files (as a whole) | 15:41 |
| daemontool | then we can reuse the rync, tar coe | 15:41 |
| daemontool | code | 15:41 |
| daemontool | but without having the agent actually installed in the VM | 15:42 |
| daemontool | like agentless with agent | 15:42 |
| daemontool | :) | 15:42 |
| daemontool | just a proposal | 15:42 |
| daemontool | fell free of course to disagree with anything | 15:42 |
| raliev | don't have any objections for now because I still have some "white spots" in understanding a whole process :) | 15:44 |
| raliev | as far as I understand, we have to create and boot a vm with bundled freezer-agent for backup, correct? | 15:45 |
| *** yuval has quit IRC | 15:45 | |
| daemontool | raliev: yes, the same VM where the agent run, the if we want to parallelize and automate | 15:50 |
| daemontool | by creating the VM from scratch | 15:50 |
| daemontool | that's cool either | 15:50 |
| daemontool | but probalby we could do that by implementing heat resources for freezer | 15:50 |
| daemontool | somehow we need to solve the scalability issues | 15:51 |
| daemontool | without implementing storage backend specific code | 15:51 |
| daemontool | is possible | 15:51 |
| daemontool | if possible | 15:52 |
| raliev | in general, that's a good idea, if we'll find a way how to speed up the process of creating snapshot of a volume and restore it - that would be perfect :) | 15:57 |
| daemontool | yep | 15:59 |
| raliev | basically this is a main issue - get a quick access to static volume content, regardless of backend type | 16:02 |
| daemontool | yes | 16:02 |
| daemontool | agreed | 16:02 |
| *** dmellado has quit IRC | 16:07 | |
| *** yangyapeng has joined #openstack-freezer | 16:12 | |
| *** yangyapeng has quit IRC | 16:17 | |
| *** dstepanenko has quit IRC | 16:21 | |
| raliev | slashme, daemontool I've updated rsync patchset, actually there are no issues with incremental backup and backup a single file. Also I've added tests and workaround if file was deleted (it'will be marked as deleted in backup meta and unlinked during restore process). if you have a chance - please review it :) | 16:25 |
| daemontool | raliev: excellent, thanks | 16:27 |
| daemontool | about that, at some point probably we need to think about the soft/hard approach we were discussing? | 16:32 |
| *** dmellado has joined #openstack-freezer | 16:48 | |
| *** dmellado has quit IRC | 16:53 | |
| *** dmellado has joined #openstack-freezer | 16:59 | |
| *** vAlexa has quit IRC | 17:39 | |
| *** yangyapeng has joined #openstack-freezer | 17:53 | |
| *** yangyapeng has quit IRC | 17:59 | |
| *** daemontool has quit IRC | 18:02 | |
| *** vAlexa has joined #openstack-freezer | 18:20 | |
| *** vAlexa_ has joined #openstack-freezer | 18:24 | |
| *** vAlexa_ has quit IRC | 18:30 | |
| *** electrofelix has quit IRC | 18:34 | |
| *** lpetrut has quit IRC | 19:16 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/freezer: Updated from global requirements https://review.openstack.org/423663 | 19:50 |
| *** yangyapeng has joined #openstack-freezer | 19:56 | |
| *** yangyapeng has quit IRC | 20:00 | |
| *** vanez has joined #openstack-freezer | 20:40 | |
| *** yangyapeng has joined #openstack-freezer | 20:57 | |
| *** yangyapeng has quit IRC | 21:02 | |
| *** lpetrut has joined #openstack-freezer | 21:10 | |
| *** NostawRm has joined #openstack-freezer | 21:16 | |
| *** catintheroof has quit IRC | 21:35 | |
| *** catintheroof has joined #openstack-freezer | 21:35 | |
| *** catintheroof has quit IRC | 21:35 | |
| *** vanez has quit IRC | 21:58 | |
| NostawRm | am i reading the docs correctly that the incremental backups only work with swift? | 22:41 |
| *** lpetrut has quit IRC | 22:44 | |
| *** yangyapeng has joined #openstack-freezer | 22:58 | |
| *** yangyapeng has quit IRC | 23:03 | |
| *** yangyapeng has joined #openstack-freezer | 23:44 | |
| *** yangyapeng has quit IRC | 23:48 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!