09:10:11 <aspiers> #startmeeting ha 09:10:12 <openstack> Meeting started Mon Dec 5 09:10:11 2016 UTC and is due to finish in 60 minutes. The chair is aspiers. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:10:13 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:10:15 <openstack> The meeting name has been set to 'ha' 09:10:21 <ddeja> o/ 09:10:28 <aspiers> ddeja, haukebruno: I guess we'll start, maybe others will join in a bit 09:10:40 <aspiers> #topic specs 09:11:03 <aspiers> so I totally forgot to submit that spec after last week's meeting so I submitted it last night 09:11:14 <aspiers> I made some more improvements to it too 09:11:26 * ddeja adds to review queue 09:11:31 <aspiers> https://review.openstack.org/#/c/406659/ 09:11:56 <aspiers> however I realised that the template does not quite have the perfect structure for it 09:12:10 <aspiers> and that is causing the tests to fail when run locally 09:12:17 <ddeja> oh 09:12:27 <aspiers> I didn't yet check why the Jenkins check is passing 09:12:44 <aspiers> actually I'm not even sure how to see the build log 09:12:53 <aspiers> since it's not linked from gerrit any more 09:12:59 <aspiers> :-/ 09:13:42 <aspiers> oh crap, looks like I uploaded an old version 09:14:11 <ddeja> OK, so I'm waiting with reviewing 09:14:14 <ddeja> :) 09:15:16 <aspiers> uploading a new one now 09:15:22 <ddeja> cool 09:16:26 <aspiers> done 09:16:53 <aspiers> let me know what you think about the new section structure 09:17:04 <ddeja> right now? 09:17:07 <aspiers> no :) 09:17:14 <aspiers> I tried to emphasise the line between the architecture and the actual implementation 09:17:32 <ddeja> that's good 09:17:39 <aspiers> to make sure each component stays compatible whilst remaining independent 09:17:47 <aspiers> no coupling between components 09:17:58 <aspiers> I wasn't too sure about the work items though 09:19:04 <beekhof> oh goody, we;re having a meeting :) 09:19:14 <aspiers> :) 09:19:16 <ddeja> \o/ 09:19:21 <aspiers> next spec for me is libvirt OCF RA 09:20:39 <ddeja> it's already in review, isn't it? 09:20:40 <aspiers> beekhof: I took the liberty of giving you the work item around the ha-guide, but I will try to help with that 09:20:58 <beekhof> ok :) 09:21:06 <aspiers> ddeja: I don't think so? 09:21:19 <beekhof> back in 10 09:22:28 <ddeja> aspiers: oh, so I must have confuse it with something else 09:22:34 <aspiers> no problem 09:26:06 <aspiers> sorry, simultaneously in my team's standup meeting 09:26:49 <ddeja> like on IRC or skype? 09:27:00 <aspiers> on VoIP (mumble) 09:27:04 <aspiers> will be finished in a sec 09:27:14 <aspiers> there is one detail missing from the compute node monitoring spec 09:27:33 <ddeja> which is? 09:27:42 <aspiers> getting the link 09:28:23 <aspiers> https://etherpad.openstack.org/p/newton-instance-ha line 135 09:28:33 <aspiers> ddeja: this was something discussed in Barcelona 09:28:49 <aspiers> and line 140-141 09:29:33 <ddeja> but this is about host recovery, not monitoring? 09:29:40 <aspiers> so the idea was to make fence_evacuate implement not just the spec but also an extra feature of being able to notify masakari 09:30:05 <aspiers> although now I can't remember why masakari can't just be modified to accept the same format 09:30:25 <aspiers> ddeja: well, it's about the message passed between monitoring and recovery 09:30:34 <ddeja> ok 09:30:46 <ddeja> by fence_evacuate you mean the resource agent? 09:30:57 <aspiers> no the fencing agent you wrote 09:31:05 * ddeja is getting lost in all names... 09:31:08 <aspiers> :) 09:31:08 <ddeja> OK 09:31:43 <ddeja> IMO we should provide minimal script 09:32:11 <ddeja> or separate script for masakari 09:32:18 <ddeja> so we can keep it simple 09:32:32 <aspiers> that's true 09:32:37 <aspiers> although they might have a lot of shared code 09:32:49 <beekhof> back 09:32:51 <aspiers> also I was wondering where is the best home for fence_evacuate.py 09:33:02 <ddeja> python script with shared base class? 09:33:04 <aspiers> beekhof: I added you to this review 09:33:10 <aspiers> ddeja: yeah maybe 09:33:42 <ddeja> I'm just against mixing two mechanisms in one fence agent 09:34:09 <ddeja> that's where my conserns came from 09:34:14 <aspiers> ok 09:36:22 <beekhof> fence what? 09:36:30 <beekhof> what does this agent do? 09:36:40 <aspiers> beekhof: notify the recovery workflow controller 09:36:53 <ddeja> POC agent that sends API call to mistral 09:36:54 <aspiers> beekhof: it's basically fence_evacuate.py 09:37:11 <aspiers> but we want to support notifying masakari too 09:37:55 <ddeja> That's too bad that there is no NTT guys, but I'd like to ask it anyway: Why? 09:38:09 <beekhof> they cant use fence_compute? 09:38:13 <ddeja> I mean, what value is added by Masakari? 09:38:23 <aspiers> ddeja: quite a lot these days, actually 09:38:27 <aspiers> ddeja: it got a lot better 09:38:44 <aspiers> ddeja: if you look in new big tent gerrit repos 09:39:00 <ddeja> OK, so it would be for, hmm, easier transition? 09:39:59 <ddeja> to be well understood: I'm not saying that Masakari is not needed. I just want to better understood what happend in Barcelona :) 09:40:22 <aspiers> ddeja: NTT could probably implement the masakari interface bit 09:41:05 <ddeja> OK 09:41:35 <aspiers> ddeja: the long-term idea would be for masakari to receive the notification and then optionally pass it on to mistral 09:41:50 <aspiers> sorry, VoIP meeting is distracting again 09:42:25 <aspiers> ddeja: so mistral would be used to implement the recovery, and masakari would implement the policy 09:42:33 <ddeja> oh, OK 09:42:41 <ddeja> what happend to congress? 09:42:47 <ddeja> as a policy engine? 09:43:00 <aspiers> that could happen in the longer term 09:43:15 <ddeja> OK 09:43:36 * ddeja starts to understand it 09:47:31 <aspiers> #topic meeting time 09:47:44 <aspiers> would people be ok with moving the meeting to another day? 09:47:50 <aspiers> Monday's are really difficult for me 09:47:58 <beekhof> yes 09:48:13 <aspiers> and it would be nice to have some time to prepare for the meeting which is not Sunday 09:48:26 <aspiers> I guess we need to keep the same time due to timezones 09:48:27 <ddeja> every day except Friday fits me (if it stays around morning in UTC) 09:48:48 <aspiers> ddeja: ok thanks 09:49:03 <aspiers> I think Wed/Thurs would be best for me 09:49:11 <aspiers> we often have meetings on Mon/Tues 09:49:19 <aspiers> beekhof: any preference 09:49:36 <beekhof> tuesday is good 09:49:38 <aspiers> I can submit a review for the day change and add everyone to review it 09:49:43 <beekhof> but wednesday can also work 09:49:51 <beekhof> thursday is out 09:51:00 <aspiers> ok 09:51:04 <aspiers> I'll submit for Wed then 09:51:08 <aspiers> thanks 09:54:54 <aspiers> #topic AOB 09:54:59 <aspiers> anything else? 09:55:07 <ddeja> nothing from my side 09:55:11 <aspiers> sorry, I am still in this other important meeting, difficult to multi-task 09:55:27 <ddeja> no problem 09:58:24 <aspiers> alright, I guess we can end if nothing else 09:58:27 <aspiers> thanks all 09:58:30 <aspiers> see you in gerrit :) 09:58:33 <aspiers> #endmeeting