aspiers | beekhof_mb: remind me why you split fence_evacuate out of fence_compute? | 00:00 |
---|---|---|
*** markvoelker has joined #openstack-ha | 00:11 | |
beekhof_mb | aspiers: because they're two very different things | 00:27 |
beekhof_mb | aspiers: and it allows for much better error handling | 00:27 |
*** yangyapeng has quit IRC | 00:27 | |
beekhof_mb | in both | 00:28 |
*** yangyapeng has joined #openstack-ha | 01:00 | |
*** yangyapeng has quit IRC | 01:04 | |
*** yangyapeng has joined #openstack-ha | 01:14 | |
*** kgaillot has quit IRC | 01:22 | |
*** vuntz has quit IRC | 01:41 | |
*** vuntz has joined #openstack-ha | 01:42 | |
*** vuntz has joined #openstack-ha | 01:42 | |
*** masber has joined #openstack-ha | 02:00 | |
*** threestrands_ has joined #openstack-ha | 02:02 | |
*** masuberu has quit IRC | 02:02 | |
*** jappleii__ has quit IRC | 02:04 | |
*** markvoelker has quit IRC | 03:13 | |
*** jmlowe has quit IRC | 03:44 | |
*** samP has joined #openstack-ha | 03:55 | |
*** ekcs has quit IRC | 05:16 | |
*** ekcs has joined #openstack-ha | 05:17 | |
*** markvoelker has joined #openstack-ha | 05:44 | |
*** markvoelker has quit IRC | 05:49 | |
*** ekcs has quit IRC | 05:55 | |
*** ekcs has joined #openstack-ha | 06:01 | |
*** zzhou_tb1 has joined #openstack-ha | 06:35 | |
*** zzhou_tb has quit IRC | 06:36 | |
*** zzhou_tb1 is now known as zzhou_tb | 06:36 | |
*** threestrands_ has quit IRC | 07:18 | |
*** zzhou_tb has quit IRC | 07:38 | |
*** zzhou_tb has joined #openstack-ha | 07:42 | |
*** markvoelker has joined #openstack-ha | 07:46 | |
*** samP has quit IRC | 07:50 | |
*** AlexeyAbashkin has joined #openstack-ha | 08:15 | |
*** markvoelker has quit IRC | 08:19 | |
*** jpena|off is now known as jpena | 08:46 | |
*** samP has joined #openstack-ha | 08:52 | |
*** zzhou_tb1 has joined #openstack-ha | 08:59 | |
*** masahito has joined #openstack-ha | 09:00 | |
*** zzhou_tb has quit IRC | 09:01 | |
*** zzhou_tb1 is now known as zzhou_tb | 09:01 | |
*** yangyapeng has quit IRC | 10:04 | |
*** yangyapeng has joined #openstack-ha | 10:04 | |
*** rossella_s has quit IRC | 10:16 | |
*** markvoelker has joined #openstack-ha | 10:16 | |
*** AlexeyAbashkin has quit IRC | 10:23 | |
*** AlexeyAbashkin has joined #openstack-ha | 10:26 | |
*** markvoelker has quit IRC | 10:50 | |
aspiers | beekhof_mb: I'm definitely not against it, and very likely in favour, just trying to understand better | 11:10 |
beekhof_mb | didnt we already have it as two separate boxes in your diagrams? | 11:11 |
aspiers | beekhof_mb: so NovaCompute / nova-compute-wait only call fence_compute, but NovaEvacuate calls fence_evacuate if it's available or fence_compute otherwise | 11:12 |
aspiers | right? | 11:12 |
beekhof_mb | makes it easier to swap fence_evacuate out for whatever the green box will be | 11:12 |
aspiers | I can't remember :) | 11:12 |
beekhof_mb | we dont have nova-compute-wait | 11:12 |
* aspiers looks at the diagram | 11:12 | |
aspiers | oh, you dropped that? | 11:12 |
beekhof_mb | long time ago | 11:12 |
aspiers | the fence_evacuate in that diagram was the one ddeja wrote for talking to mistral | 11:13 |
aspiers | https://aspiers.github.io/openstack-day-israel-2017-compute-ha/#/nova-host-alerter | 11:13 |
aspiers | that's completely different to your fence_evacuate | 11:14 |
beekhof_mb | in any case, two agents makes it easy to trigger something that isn't as dumb as our current fence_evacuate | 11:14 |
beekhof_mb | while keeping the important bits | 11:14 |
aspiers | hrm, it seems the diagram is missing an arrow from NovaCompute to fence_compute :-/ | 11:15 |
*** ushkalim has joined #openstack-ha | 11:15 | |
aspiers | I'm pretty sure I'm still missing one key piece of architectural logic | 11:15 |
* aspiers waits for the penny to drop | 11:15 | |
beekhof_mb | for us its not fence_evacuate but right | 11:15 |
beekhof_mb | ok | 11:15 |
beekhof_mb | so in that diagram you have fence_evacuate and fence_compute | 11:16 |
beekhof_mb | one is red. the other yellow | 11:16 |
aspiers | right | 11:16 |
beekhof_mb | much easier to retire the red one if its not in the same file as the yellow one. no? | 11:16 |
aspiers | yes, but it never was in the same file | 11:17 |
beekhof_mb | yes it was | 11:17 |
aspiers | https://github.com/gryf/mistral-evacuate/blob/master/fence_evacuate.py | 11:17 |
aspiers | it was that one | 11:17 |
aspiers | entirely different thing | 11:17 |
aspiers | only talks to mistral | 11:17 |
beekhof_mb | and much easier to swap in a mitral talking fence_evacuate for our shit nova one if its not in the same file as fence_compute | 11:17 |
beekhof_mb | s/mistral/whatever/ | 11:18 |
beekhof_mb | the concept is the same though | 11:18 |
aspiers | is it? | 11:18 |
beekhof_mb | an agent that actually does the evacuation | 11:18 |
beekhof_mb | or triggers it | 11:18 |
beekhof_mb | it should not be in the same agent/file that is making nova safe before the evac can be performed | 11:19 |
aspiers | well, the mechanism for choosing a different evac method was supposed to be nova-host-alerter | 11:19 |
beekhof_mb | fence_compute <-- make nova safe | 11:19 |
aspiers | but yes I agree on the value of separation, absolutely | 11:19 |
beekhof_mb | sure, so s/fence_evacuate/nova-host-alerter/ in the future | 11:20 |
aspiers | OK, think I got it now | 11:20 |
aspiers | ... in a woolly Tuesday morning without coffee sort of way :) | 11:21 |
beekhof_mb | :) | 11:21 |
beekhof_mb | before, fence_compute == make nova safe + trigger evacuation | 11:21 |
beekhof_mb | before, fence_compute == make nova safe + trigger evacuation in a very specific way | 11:21 |
beekhof_mb | and we want that way to change | 11:21 |
beekhof_mb | so pull out that bit so we can ditch it more easily later and still keep the generically useful bit | 11:22 |
aspiers | by "make nova safe" you mean fence + mark host down, right? | 11:22 |
beekhof_mb | power fence happens elsewhere | 11:22 |
beekhof_mb | but yes, mark down and disable | 11:22 |
aspiers | ah yeah, of course | 11:22 |
beekhof_mb | btw. https://github.com/beekhof/galera-operator/blob/master/apps/galera/deployment.yaml | 11:22 |
aspiers | multi-level fencing | 11:22 |
beekhof_mb | galera managed natively in k8 | 11:23 |
beekhof_mb | i'll announce in due course | 11:23 |
aspiers | oh, well done :) | 11:23 |
aspiers | you've been working on that a while now | 11:23 |
beekhof_mb | thinking about it for a while | 11:23 |
beekhof_mb | working only recently | 11:23 |
aspiers | "ReplicatedStatefulSet" huh | 11:23 |
aspiers | interesting | 11:23 |
beekhof_mb | least worst name i could come up with | 11:24 |
beekhof_mb | short form... rss :) | 11:24 |
aspiers | haha nooooo | 11:25 |
aspiers | that's taken already :) | 11:25 |
aspiers | guessing you're not coming to Dublin? | 11:25 |
beekhof_mb | nope | 11:25 |
aspiers | but probably Vancouver? | 11:25 |
beekhof_mb | and yes, i'm aware its in use outside of k8 | 11:25 |
beekhof_mb | unlikely but stranger things have happened | 11:26 |
aspiers | otp | 11:28 |
beekhof_mb | otp? | 11:35 |
*** rossella_s has joined #openstack-ha | 11:44 | |
aspiers | on the phone | 11:57 |
*** masahito has quit IRC | 12:06 | |
*** rossella_s has quit IRC | 12:15 | |
*** rossella_s has joined #openstack-ha | 12:19 | |
*** pcaruana has joined #openstack-ha | 12:33 | |
*** rossella_s has quit IRC | 12:35 | |
*** jpena is now known as jpena|lunch | 12:35 | |
*** rossella_s has joined #openstack-ha | 12:37 | |
*** rossella_s has quit IRC | 12:43 | |
*** rossella_s has joined #openstack-ha | 12:45 | |
*** markvoelker has joined #openstack-ha | 12:47 | |
*** yangyapeng has quit IRC | 12:56 | |
*** rossella_s has quit IRC | 13:03 | |
*** rossella_s has joined #openstack-ha | 13:07 | |
*** markvoelker has quit IRC | 13:20 | |
*** rossella_s has quit IRC | 13:21 | |
*** yangyapeng has joined #openstack-ha | 13:22 | |
*** rossella_s has joined #openstack-ha | 13:22 | |
*** bandini has quit IRC | 13:28 | |
*** markvoelker has joined #openstack-ha | 13:33 | |
*** bandini has joined #openstack-ha | 13:35 | |
*** jpena|lunch is now known as jpena | 13:39 | |
*** rossella_s has quit IRC | 13:40 | |
*** rossella_s has joined #openstack-ha | 13:42 | |
*** samP has quit IRC | 13:47 | |
*** samP has joined #openstack-ha | 13:50 | |
openstackgerrit | Merged openstack/openstack-resource-agents master: NovaEvacuate: Correctly handle stopped hypervisors https://review.openstack.org/428076 | 13:51 |
openstackgerrit | Merged openstack/openstack-resource-agents master: NovaEvacuate: Support the new split-out IHA fence agents with backwards compatibility https://review.openstack.org/503617 | 13:51 |
*** cleong has joined #openstack-ha | 13:52 | |
openstackgerrit | Merged openstack/openstack-resource-agents master: neutron-ha-tool: do not replicate dhcp https://review.openstack.org/463320 | 13:53 |
*** masahito has joined #openstack-ha | 14:00 | |
*** masahito has quit IRC | 14:05 | |
*** rossella_s has quit IRC | 14:07 | |
*** rossella_s has joined #openstack-ha | 14:09 | |
*** rossella_s has quit IRC | 14:15 | |
*** rossella_s has joined #openstack-ha | 14:17 | |
*** jmlowe has joined #openstack-ha | 14:27 | |
*** kgaillot has joined #openstack-ha | 14:35 | |
*** rossella_s has quit IRC | 14:59 | |
*** rossella_s has joined #openstack-ha | 15:03 | |
openstackgerrit | Merged openstack/openstack-resource-agents master: NovaEvacuate: Allow debug logging to be turned on easily https://review.openstack.org/334737 | 15:32 |
*** rossella_s has quit IRC | 15:36 | |
*** ushkalim has quit IRC | 16:13 | |
*** rossella_s has joined #openstack-ha | 16:23 | |
*** jpena is now known as jpena|off | 16:37 | |
*** jpena|off is now known as jpena | 16:41 | |
*** AlexeyAbashkin has quit IRC | 17:11 | |
*** openstackgerrit has quit IRC | 17:33 | |
*** ekcs has quit IRC | 17:52 | |
*** cleong has quit IRC | 18:33 | |
*** AlexeyAbashkin has joined #openstack-ha | 18:43 | |
*** jpena is now known as jpena|off | 18:45 | |
*** ekcs has joined #openstack-ha | 18:47 | |
*** AlexeyAbashkin has quit IRC | 18:48 | |
*** pcaruana has quit IRC | 19:50 | |
*** threestrands_ has joined #openstack-ha | 21:18 | |
*** rossella_s has quit IRC | 21:53 | |
*** rossella_s has joined #openstack-ha | 21:54 | |
*** rossella_s has quit IRC | 22:04 | |
*** rossella_s has joined #openstack-ha | 22:08 | |
*** rossella_s has quit IRC | 22:52 | |
*** rossella_s has joined #openstack-ha | 22:55 | |
*** rossella_s has quit IRC | 23:00 | |
*** rossella_s has joined #openstack-ha | 23:02 | |
*** jmlowe_ has joined #openstack-ha | 23:04 | |
*** jmlowe has quit IRC | 23:05 | |
*** rossella_s has quit IRC | 23:08 | |
*** rossella_s has joined #openstack-ha | 23:12 | |
*** sticker has joined #openstack-ha | 23:38 | |
*** rossella_s has quit IRC | 23:51 | |
*** rossella_s has joined #openstack-ha | 23:54 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!