| *** cknight has quit IRC | 00:00 | |
| *** masahito has joined #openstack-ha | 00:03 | |
| *** masahito has quit IRC | 00:42 | |
| *** masahito has joined #openstack-ha | 00:43 | |
| *** dgurtner has quit IRC | 00:48 | |
| *** kgaillot has quit IRC | 01:10 | |
| *** markvoelker has quit IRC | 01:23 | |
| *** mancdaz has quit IRC | 01:39 | |
| *** mancdaz has joined #openstack-ha | 01:40 | |
| *** cknight has joined #openstack-ha | 02:04 | |
| *** cknight has quit IRC | 03:43 | |
| *** masahito has quit IRC | 04:00 | |
| *** cknight has joined #openstack-ha | 04:02 | |
| *** masahito has joined #openstack-ha | 04:43 | |
| *** cknight has quit IRC | 05:01 | |
| *** ianbrown has joined #openstack-ha | 05:14 | |
| *** ianbrown has quit IRC | 05:21 | |
| *** markvoelker has joined #openstack-ha | 05:25 | |
| *** markvoelker_ has joined #openstack-ha | 05:28 | |
| *** markvoelker has quit IRC | 05:30 | |
| *** markvoelker has joined #openstack-ha | 05:31 | |
| *** markvoelker_ has quit IRC | 05:34 | |
| *** markvoelker_ has joined #openstack-ha | 05:38 | |
| *** markvoelker has quit IRC | 05:38 | |
| *** markvoelker_ has quit IRC | 05:51 | |
| *** markvoelker has joined #openstack-ha | 06:50 | |
| *** markvoelker has quit IRC | 06:55 | |
| *** mjura has joined #openstack-ha | 07:20 | |
| *** zzhou has quit IRC | 07:45 | |
| *** zhilong has quit IRC | 07:47 | |
| *** markvoelker has joined #openstack-ha | 08:25 | |
| *** aspiers has joined #openstack-ha | 08:26 | |
| *** jpena has joined #openstack-ha | 08:27 | |
| *** mobaer has joined #openstack-ha | 08:30 | |
| *** markvoelker has quit IRC | 08:30 | |
| *** dgurtner has joined #openstack-ha | 08:49 | |
| *** dgurtner has quit IRC | 08:49 | |
| *** dgurtner has joined #openstack-ha | 08:49 | |
| *** masahito has quit IRC | 09:34 | |
| *** masahito has joined #openstack-ha | 09:37 | |
| *** masahito has quit IRC | 09:41 | |
| *** markvoelker has joined #openstack-ha | 10:26 | |
| *** markvoelker has quit IRC | 10:31 | |
| *** dgurtner has quit IRC | 10:47 | |
| *** dgurtner has joined #openstack-ha | 10:48 | |
| *** markvoelker has joined #openstack-ha | 12:27 | |
| *** markvoelker has quit IRC | 12:32 | |
| *** openstackgerrit has quit IRC | 12:32 | |
| *** openstackgerrit has joined #openstack-ha | 12:33 | |
| *** openstackgerrit has quit IRC | 13:17 | |
| *** openstackgerrit has joined #openstack-ha | 13:18 | |
| *** smoriya_ has quit IRC | 13:20 | |
| aspiers | ddeja, _gryf: so it seems the Congress project has a use case covering the same thing you are working on with Mistral | 13:20 |
|---|---|---|
| aspiers | ddeja, _gryf: to me, it would make most sense if we keep the execution part in Mistral, but then use Congress to trigger evacuation based on whatever policy the cloud operator chooses (e.g. some kind of SLA for pets) | 13:21 |
| aspiers | Congress does not yet integrate with Mistral however. It was planned for liberty but apparently didn't happen | 13:22 |
| aspiers | https://wiki.openstack.org/wiki/PolicyGuidedFulfillment#Proposed_Approach | 13:22 |
| aspiers | https://wiki.openstack.org/wiki/PolicyGuidedFulfillmentLibertyPlanning#Congress | 13:22 |
| aspiers | the use case is described here: https://docs.google.com/document/d/1ExDmT06vDZjzOPePYBqojMRfXodvsk0R8nRkX-zrkSw/edit#heading=h.8eofb0gqkota | 13:22 |
| aspiers | https://wiki.openstack.org/wiki/PolicyGuidedFulfillmentLibertyPlanning_Remediation | 13:23 |
| *** mobaer has quit IRC | 13:24 | |
| aspiers | apparently there is no blueprint yet | 13:24 |
| *** mobaer has joined #openstack-ha | 13:27 | |
| *** markvoelker has joined #openstack-ha | 13:28 | |
| *** markvoelker has quit IRC | 13:32 | |
| _gryf | aspiers, what is Congress again? Some kind of decision making service? Isn't the pacemaker the one who have all the knowledge of the dead host, so it's the one who is going to trigger something anyway? | 13:48 |
| *** jpena is now known as jpena|lunch | 13:48 | |
| *** markvoelker has joined #openstack-ha | 13:50 | |
| *** markvoelker_ has joined #openstack-ha | 13:52 | |
| *** markvoelker has quit IRC | 13:54 | |
| _gryf | aspiers, I think, you were pointed at "Automatic Evacuation on Host Failure" (voted with +3) | 14:15 |
| _gryf | I think, that within OpenStack people doesn't want to go different way than OS services, i.e. when I see ceilometer as a source of state of the host it make my eyes water | 14:19 |
| aspiers | _gryf: https://wiki.openstack.org/wiki/Congress | 14:28 |
| aspiers | _gryf: yes, pacemaker has the knowledge of the dead host first, but no it's not the one to trigger anything | 14:28 |
| aspiers | pacemaker simply marks the host as dead and makes sure that nova knows it's dead | 14:29 |
| aspiers | the subsequent remedial action is policy-based, and really depends on what the cloud operator and customer want the desired behaviour to be | 14:29 |
| aspiers | e.g. the operator might want to charge extra for this service | 14:30 |
| aspiers | so it doesn't evacuate all hosts | 14:30 |
| aspiers | only ones which contain pets from customers which have paid extra | 14:30 |
| aspiers | or it could be done per availability zone | 14:30 |
| aspiers | but we don't know that, because policies will be different for each cloud | 14:30 |
| aspiers | that's why it would be really nice to involve Congress | 14:31 |
| aspiers | since that's exactly the problem it solves | 14:31 |
| aspiers | _gryf: we certainly don't need to use ceilometer for this. Congress can take data-sources from anywhere | 14:31 |
| _gryf | aspiers, well, I don;t know much about congress, in our solution it is as simple as marking the flavour OR the vm itself via extra spec or tagging vm. Is it possible via congress? | 14:35 |
| aspiers | yes | 14:36 |
| _gryf | ok cool. | 14:36 |
| aspiers | _gryf: I don't know much about congress either, but my colleague just gave a presentation on it, and it seems like it would be the perfect complement to your mistral work | 14:36 |
| aspiers | I think they would tie together perfectly | 14:36 |
| aspiers | so mistral would be in charge of doing the work, and congress would decide when to do it | 14:36 |
| _gryf | I'm wondering how fast would be reaction from host outage to triggering the evac process | 14:36 |
| aspiers | yes, good question | 14:37 |
| _gryf | since the goal would not "automatic" in the "automatic evacuation" title, but it have to be fast, fast, fast :D | 14:38 |
| _gryf | automatic too, of course, but not only one | 14:38 |
| aspiers | right | 14:40 |
| aspiers | anyway, I thought I should mention this even though it is probably not a topic for immediate action | 14:41 |
| aspiers | but we should at least keep some eyes on Congress :) | 14:41 |
| aspiers | and watch for when they add Mistral integration | 14:41 |
| aspiers | or maybe we can even help with that | 14:41 |
| _gryf | yes, definitely. let's consider all the options we have. | 14:43 |
| *** kgaillot has joined #openstack-ha | 14:52 | |
| *** sigmavirus24_awa is now known as sigmavirus24 | 14:57 | |
| *** jpena|lunch is now known as jpena | 15:00 | |
| *** cknight has joined #openstack-ha | 15:17 | |
| *** markvoelker_ has quit IRC | 15:25 | |
| *** mancdaz has quit IRC | 15:31 | |
| *** mancdaz has joined #openstack-ha | 15:31 | |
| *** markvoelker has joined #openstack-ha | 15:55 | |
| *** cknight has quit IRC | 16:20 | |
| *** mjura has quit IRC | 16:45 | |
| *** holser_ has quit IRC | 17:00 | |
| *** holser_ has joined #openstack-ha | 17:00 | |
| *** jpena has quit IRC | 17:33 | |
| *** aspiers has quit IRC | 17:50 | |
| *** holser_ has quit IRC | 18:31 | |
| *** holser_ has joined #openstack-ha | 18:33 | |
| *** dgurtner has quit IRC | 19:18 | |
| *** openstackgerrit has quit IRC | 20:47 | |
| *** openstackgerrit has joined #openstack-ha | 20:48 | |
| *** mancdaz has quit IRC | 20:58 | |
| *** mancdaz has joined #openstack-ha | 21:01 | |
| *** aspiers has joined #openstack-ha | 21:13 | |
| *** dbel has joined #openstack-ha | 22:06 | |
| *** mobaer has quit IRC | 22:28 | |
| *** dbel has quit IRC | 22:44 | |
| *** aspiers has quit IRC | 23:10 | |
| *** sigmavirus24 is now known as sigmavirus24_awa | 23:57 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!