*** markvoelker_ has quit IRC | 00:10 | |
*** cknight has quit IRC | 00:14 | |
*** markvoelker has joined #openstack-ha | 01:10 | |
*** markvoelker has quit IRC | 01:15 | |
*** beekhof has quit IRC | 01:24 | |
*** zzhou has quit IRC | 01:45 | |
*** zhilong has quit IRC | 01:45 | |
*** zzhou has joined #openstack-ha | 01:50 | |
*** zhilong has joined #openstack-ha | 01:57 | |
*** ianbrown has quit IRC | 02:07 | |
*** ianbrown has joined #openstack-ha | 02:08 | |
*** ianbrown has quit IRC | 03:02 | |
*** ianbrown has joined #openstack-ha | 03:02 | |
*** markvoelker has joined #openstack-ha | 03:11 | |
*** markvoelker has quit IRC | 03:16 | |
*** cknight has joined #openstack-ha | 03:18 | |
*** ianbrown has quit IRC | 03:25 | |
*** ianbrown has joined #openstack-ha | 03:25 | |
*** ianbrown has quit IRC | 03:36 | |
*** ianbrown has joined #openstack-ha | 03:36 | |
*** ianbrown has quit IRC | 03:41 | |
*** ianbrown has joined #openstack-ha | 03:41 | |
*** mihgen has quit IRC | 04:36 | |
*** mihgen has joined #openstack-ha | 04:38 | |
*** beekhof has joined #openstack-ha | 04:52 | |
*** markvoelker has joined #openstack-ha | 05:12 | |
*** ianbrown has quit IRC | 05:17 | |
*** markvoelker has quit IRC | 05:17 | |
*** ianbrown has joined #openstack-ha | 05:17 | |
*** cknight has quit IRC | 05:40 | |
*** gomarivera has quit IRC | 06:04 | |
*** zzhou has quit IRC | 06:04 | |
*** masahito has joined #openstack-ha | 06:24 | |
*** masahito has quit IRC | 06:25 | |
*** masahito has joined #openstack-ha | 06:35 | |
*** mjura has joined #openstack-ha | 06:57 | |
*** openstackgerrit has joined #openstack-ha | 07:10 | |
*** markvoelker has joined #openstack-ha | 07:13 | |
*** masahito has quit IRC | 07:15 | |
*** markvoelker has quit IRC | 07:17 | |
*** masahito has joined #openstack-ha | 07:19 | |
*** rmart04 has joined #openstack-ha | 07:26 | |
openstackgerrit | Merged openstack/ha-guide: Updated from openstack-manuals https://review.openstack.org/257164 | 07:30 |
---|---|---|
*** vuntz has quit IRC | 07:52 | |
*** vuntz has joined #openstack-ha | 07:54 | |
*** vuntz has quit IRC | 07:54 | |
*** vuntz has joined #openstack-ha | 07:54 | |
*** mobaer has joined #openstack-ha | 08:16 | |
*** mobaer has quit IRC | 08:23 | |
*** masahito has quit IRC | 08:23 | |
*** masahito has joined #openstack-ha | 08:24 | |
*** jpena has joined #openstack-ha | 08:31 | |
*** kazuIchikawa has joined #openstack-ha | 08:41 | |
aspiers | morning :) | 08:53 |
aspiers | (or evening) | 08:53 |
masahito | good morning and evening :) | 08:55 |
*** vuntz has quit IRC | 09:00 | |
*** vuntz has joined #openstack-ha | 09:00 | |
*** vuntz has quit IRC | 09:00 | |
*** vuntz has joined #openstack-ha | 09:00 | |
*** markvoelker has joined #openstack-ha | 09:14 | |
*** markvoelker has quit IRC | 09:18 | |
*** rmart04_ has joined #openstack-ha | 09:47 | |
*** rmart04 has quit IRC | 09:48 | |
*** rmart04_ is now known as rmart04 | 09:48 | |
*** bandini has quit IRC | 09:56 | |
*** bandini has joined #openstack-ha | 09:56 | |
ddeja | aspiers: I wanted to bring one topic on meeting, but we've run out of time | 10:04 |
ddeja | So we can add this to agenda for next week? | 10:04 |
aspiers | yep? | 10:04 |
aspiers | sure | 10:04 |
ddeja | Ok | 10:04 |
aspiers | feel free to edit https://wiki.openstack.org/wiki/Meetings/HATeamMeeting | 10:04 |
ddeja | OK | 10:04 |
ddeja | Just short description: Let's assume that we have 6 computes | 10:05 |
ddeja | load for about 75% | 10:05 |
ddeja | and 3 of them dies - which VMs we want to evacuate | 10:05 |
ddeja | and how to choose pets ;) | 10:05 |
aspiers | that's where I see Congress being really helpful | 10:06 |
aspiers | there is no single right approach across all clouuds | 10:06 |
aspiers | it needs to be policy-based | 10:06 |
ddeja | I know, but I'm wondering about another problem | 10:06 |
aspiers | Congress gives the opportunity to control which hosts/VMs Mistral evacuates first | 10:07 |
ddeja | becouse right now we will fire up 3 evacuation processes | 10:07 |
ddeja | and we have to find a way to synchronize those evacuations | 10:07 |
ddeja | because we may want to evacuate some from host A, some from HostB and so one | 10:08 |
ddeja | having in mind limited resources | 10:08 |
masahito | ddeja: IMO, we have 2 approach. | 10:08 |
ddeja | and maybe killing some instances on alive hosts, to make place for VIVs (Very Important VMs) :) | 10:08 |
aspiers | ddeja: yes, I agree this is a challenge | 10:09 |
aspiers | although to some extent OpenStack (especially nova-scheduler) should handle the synchronization for us | 10:09 |
masahito | If a cloud is allow over commit, we can move all VMs. If not allowed, we only choose priority VMs. | 10:09 |
aspiers | masahito: priority VMs should always be resurrected first IMHO | 10:10 |
ddeja | aspiers: Nova-scheduler wouldn't handle, since we avacuate all VMs one-by-one | 10:10 |
aspiers | ddeja: why is that a problem? | 10:11 |
aspiers | as long as markdown API can be used in parallel | 10:11 |
ddeja | Let's assume that on compute 1 we have VMS with priority 1,2,3 and on compute 2 1,3,3 | 10:11 |
aspiers | ok | 10:11 |
ddeja | both dies, and on compute 3 we have place for 3 vms | 10:12 |
ddeja | so if evacuate form ompute 2 will be fast, it vm's may be evacuated faster then those from compute 1 | 10:12 |
aspiers | I see the proble | 10:12 |
aspiers | problem | 10:12 |
aspiers | but there is a tradeoff | 10:12 |
aspiers | e.g. what if compute1 dies 5 seconds after compute2? | 10:13 |
aspiers | or 10 seconds? 20 seconds? | 10:13 |
ddeja | that's what me and _gryf hit todays morning ;) | 10:13 |
aspiers | we can aggregate resurrection tasks into time buckets if we want | 10:13 |
aspiers | but the bigger the bucket, the slower the resurrection | 10:13 |
aspiers | and the smaller the bucket, the greater the risk of priority inversion | 10:14 |
ddeja | maybe it will be good idea to prioritize over whole cloud | 10:14 |
aspiers | (BTW I mean something different to classic kernel-level priority inversion) | 10:14 |
ddeja | so in case of failure we can even kill some low priority VM | 10:14 |
ddeja | aspiers: me too ;) | 10:14 |
aspiers | ddeja: yes, Congress could definitely have policies about VM priority | 10:15 |
aspiers | ddeja: I guess you could hardcode such a policy into Mistral for now | 10:15 |
aspiers | and then we could move it into Congress later | 10:16 |
aspiers | sounds like a longer-term goal to get this right | 10:16 |
ddeja | aspiers: agree | 10:16 |
ddeja | hmm, so I don't know if there is still need to put it on agenda for next week ;) | 10:17 |
masahito | aspiers: ddeja: agree, from Congress dev side ;) | 10:17 |
aspiers | ddeja: hehe :) | 10:17 |
aspiers | ddeja: well you can at least give a status update summarising this conversation ;-) | 10:17 |
ddeja | OK | 10:18 |
aspiers | masahito: :) | 10:18 |
aspiers | btw do you know if there is a blueprint for mistral HA yet? | 10:18 |
aspiers | and same question for Congress HA | 10:18 |
ddeja | aspiers: I don't think so | 10:18 |
masahito | aspiers: what's for? | 10:19 |
ddeja | on last Mistral meeting they've said it's still not started | 10:19 |
ddeja | And guy how's supossed to do it was working on critical bug, so I don't know if it already started or not | 10:20 |
ddeja | Will know after todays meeting (and probalby offer my help, if it's still not started) | 10:20 |
masahito | oops, your question would be for Congress HA. | 10:20 |
masahito | Congress has already had its HA feature. | 10:21 |
aspiers | ddeja: ok thanks | 10:21 |
aspiers | masahito: ah ok, nice | 10:21 |
masahito | And for Mitaka, we try to separate a process to non-state process and state full process like other project do. | 10:23 |
aspiers | makes sense | 10:35 |
*** mihgen has quit IRC | 11:00 | |
*** masahito has quit IRC | 11:02 | |
*** rmart04 has quit IRC | 11:10 | |
*** masahito has joined #openstack-ha | 11:11 | |
*** masahito has quit IRC | 11:16 | |
*** dgurtner has joined #openstack-ha | 12:09 | |
*** dgurtner has joined #openstack-ha | 12:09 | |
*** markvoelker has joined #openstack-ha | 12:45 | |
*** mobaer has joined #openstack-ha | 12:48 | |
*** markvoelker has quit IRC | 12:49 | |
*** smoriya_ has quit IRC | 13:00 | |
*** mobaer has quit IRC | 13:09 | |
*** dgurtner has quit IRC | 13:14 | |
*** mobaer has joined #openstack-ha | 13:21 | |
*** jpena is now known as jpena|lunch | 14:00 | |
*** fabbione has quit IRC | 14:03 | |
*** fabbione has joined #openstack-ha | 14:05 | |
*** mobaer has quit IRC | 14:11 | |
*** cknight has joined #openstack-ha | 14:22 | |
*** kgaillot has joined #openstack-ha | 14:34 | |
*** dgurtner has joined #openstack-ha | 14:41 | |
*** dgurtner has joined #openstack-ha | 14:41 | |
*** markvoelker has joined #openstack-ha | 14:43 | |
*** sigmavirus24_awa is now known as sigmavirus24 | 15:05 | |
*** jpena|lunch is now known as jpena | 15:08 | |
*** mihgen has joined #openstack-ha | 16:16 | |
*** dgurtner has quit IRC | 16:23 | |
*** mjura has quit IRC | 16:35 | |
*** mobaer has joined #openstack-ha | 17:18 | |
*** jpena has quit IRC | 17:32 | |
*** dgurtner has joined #openstack-ha | 17:58 | |
*** dgurtner has joined #openstack-ha | 17:58 | |
*** zhilong has quit IRC | 18:01 | |
*** zhilong has joined #openstack-ha | 18:02 | |
*** dgurtner has quit IRC | 18:33 | |
*** gwei3 has joined #openstack-ha | 19:32 | |
*** gwei3 has quit IRC | 19:43 | |
*** gwei3 has joined #openstack-ha | 19:43 | |
*** gwei3 has quit IRC | 19:45 | |
*** geguileo has quit IRC | 20:29 | |
*** cknight has quit IRC | 21:44 | |
*** cknight has joined #openstack-ha | 22:13 | |
openstackgerrit | KATO Tomoyuki proposed openstack/ha-guide: Update the overview of keepalived and pacemaker HA https://review.openstack.org/247887 | 22:16 |
*** cknight has quit IRC | 22:31 | |
*** sigmavirus24 is now known as sigmavirus24_awa | 23:22 | |
*** kazuIchikawa has quit IRC | 23:35 | |
*** mobaer has quit IRC | 23:51 | |
*** smoriya_ has joined #openstack-ha | 23:52 | |
*** markvoelker has quit IRC | 23:55 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!