08:03:19 #startmeeting ha 08:03:20 Meeting started Mon Jul 25 08:03:19 2016 UTC and is due to finish in 60 minutes. The chair is aspiers. Information about MeetBot at http://wiki.debian.org/MeetBot. 08:03:22 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 08:03:24 The meeting name has been set to 'ha' 08:03:25 hi everyone 08:03:31 hi all 08:03:47 #topic Current status (progress, issues, roadblocks, further plans) 08:03:59 let's do a quick round as usual - I'll start 08:04:14 I finished setting up the specs repo 08:04:24 thank you 08:04:26 so it is ready to use now 08:04:29 sorry it took so long 08:04:51 next thing will be to start submitting specs 08:04:52 I will send the spec today 08:04:56 cool, thanks! 08:05:05 Oh.. do we need BPs 08:05:07 ? 08:05:08 also I reviewed the cross-project spec 08:05:20 I don't think we need them 08:05:30 unless you can think of a reason? 08:05:44 specs should be enough 08:05:51 #info openstack-resource-agents-specs is ready for submissions 08:05:58 I can think any reason, so , 08:06:07 no BPs for specs 08:06:13 OK :) 08:06:27 #info we won't use Blueprints (at least until we can think of a good reason to) 08:07:06 here is the cross-project spec https://review.openstack.org/#/c/257809/ 08:07:17 it has been open for a long time I think 08:07:31 and currently it doesn't provide much value in addition to the user story 08:07:42 but in the future it can link to all the individual specs 08:08:11 it does talk about implementation details a bit 08:08:18 aspiers: Are we using this as a reference? 08:08:18 but it's mostly wrong 08:08:33 e.g. suggesting tooz can replace fencing 08:09:12 samP: I think we need to either fix the x-p spec, or remove it 08:09:24 samP: and I think the workflow the PWG requested is to fix it 08:09:38 it can talk about implementation details / architecture at a high level, which might be helpful 08:10:27 other than that, I have been busy catching up from holiday backlog 08:10:39 aspiers: originally, Timofey Durakov had a different way to implement the Instance HA 08:10:51 and I also started to discuss with beekhof about his blog post 08:11:03 samP: oh? 08:12:22 well, we can discuss after the status round 08:12:34 who wants to go next? 08:12:35 aspiers: No I mean, use Ceilometer+Mistral. which is fine. 08:12:50 oh, yeah 08:12:53 Ok, from my side, 08:13:00 the ceilometer discussion is ongoing in the spec review ... 08:13:44 Im working on 2 specs VM monitoring and VM recovery. Hope fully I will push them today and tomorrow. 08:14:04 great! 08:14:21 thatz all from my side. 08:14:29 thanks 08:14:38 anyone else? 08:15:23 nothing from my side 08:15:50 from me too 08:15:55 ok 08:16:03 #topic HA guide 08:16:16 so there was some discussion on the list about the future of the guide 08:16:38 could you please update or give some ref 08:16:49 let me try to find 08:17:12 aspiers: thanks 08:17:13 hmm, I was wrong - it was a private thread 08:17:19 but I can summarise here 08:17:20 ah.. 08:17:40 Lana Brindley said it was quite out of date (I'm not sure if this is entirely true) and they were thinking about deprecating 08:17:57 since bogdando is no longer able to look after it 08:18:06 the good news is beekhof volunteered 08:18:10 and I will also try to help him 08:18:16 great! 08:18:25 but of course any help is welcome 08:18:39 I can help 08:18:45 #info beekhof is taking over maintenance of the HA guide from bogdando; all help is still welcome 08:19:50 Lana said she'd make an announcement but I don't think it happened yet 08:20:06 #topic AOB 08:20:27 rsjethani: so we can talk about consul if you want? 08:20:33 ya sure 08:20:40 #topic Consul 08:20:47 I don't know much about it 08:20:55 I was just going throughit: https://www.consul.io/intro/ 08:21:04 right 08:21:20 it doesn't look like it can automate recovery 08:21:33 or do fencing 08:21:46 I think it's probably mainly for active/active services where fencing is not required 08:21:51 ya that was my first guess too 08:21:54 and for non-HA services 08:22:05 but ICBW 08:22:07 still i thought to sk you guys 08:22:13 sure 08:22:22 please feel free to investigate more and confirm with a Consul expert 08:22:34 ok 08:22:35 if you do, it would be great if you could let us know the outcome 08:22:41 it's always good to be aware of all the options 08:22:50 they do have the concept of watchers 08:23:17 and monitors which may be used for recovery etc. 08:23:24 there's Serf too 08:23:43 ya consul is a superset fo serf 08:23:56 it uses serf only 08:24:01 but if there is no native support for fencing then it's not a full solution for HA cluster management 08:24:18 #topic AOB 08:24:20 as aspiers mentioned above, I cant find any thing related to fence or auto recovery. maybe need to dig more 08:24:26 yep 08:24:31 anything else? 08:24:39 ya i will look further 08:24:44 I will attend to PWG mid 08:24:53 at 8/22 inNY 08:24:59 oh - probably bad news about Barcelona. the chairman of the OpenStack board asked on our behalf if there could be an HA track 08:25:09 but he didn't get any reply :-( 08:25:24 samP: cool! 08:25:32 cause I will go to ops meetup 08:25:46 samP: maybe you could look out for any of the Foundation people there, and pester them about an HA track? 08:26:15 aspiers: sure 08:27:00 aspiers: I will try. 08:27:06 thanks :) 08:27:16 I will keep trying to 08:27:17 aspiers: may be we can get some help from ops side too 08:27:22 yes, great idea 08:27:31 I will mail openstack-operators and -dev 08:27:46 gret, that will be really helpful 08:27:47 #action aspiers to email openstack-{operators,dev} about Barcelona HA track 08:28:36 any other topics? 08:28:44 anyone using pcs on ubuntu? 08:28:51 Not from my side 08:28:58 rsjethani: use crmsh ;-) 08:29:09 use crmsh 08:29:21 https://crmsh.github.io/ 08:29:22 ok :) 08:30:17 ok, last call for other topics :) 08:30:28 otherwise we can end nice and early :) 08:30:43 I am done 08:31:05 thank you all... 08:31:16 ok thanks everyone! 08:31:18 #endmeeting