18:01:10 <krtaylor> #startmeeting third-party
18:01:11 <openstack> Meeting started Mon Dec 29 18:01:10 2014 UTC and is due to finish in 60 minutes.  The chair is krtaylor. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:01:12 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
18:01:15 <openstack> The meeting name has been set to 'third_party'
18:01:26 <krtaylor> anyone here for third-party-ci meeting?
18:01:34 <ja_> i yam
18:01:43 <krtaylor> hi ja_
18:01:58 <krtaylor> this ought to be interesting  :)
18:02:15 <krtaylor> might be really short
18:02:33 <krtaylor> anyone else here?
18:02:36 <ja_> usually people are ok with very short mtgs
18:02:52 * krtaylor wouldn't mind a short meeting
18:03:41 <ja_> I think the real key is to have everyone leave thinking that it was no longer than necessary, and useful while it lasted
18:03:57 <krtaylor> well, lets go through the motions, there may be lurkers that will jump in, I'll skip sections with nothing on the agenda
18:04:23 <krtaylor> #topic Welcome & Reminder of OpenStack Mission
18:04:31 <krtaylor> #info The OpenStack Open Source Cloud Mission: to produce the ubiquitous Open Source Cloud Computing platform that will meet the needs of public and private clouds regardless of size, by being simple to implement and massively scalable.
18:04:52 <krtaylor> and the agenda:
18:04:56 <krtaylor> #link https://wiki.openstack.org/wiki/Meetings/ThirdParty#12.2F29.2F14
18:05:02 <krtaylor> not much there
18:05:42 <krtaylor> ok, we'll skip to program items
18:05:45 <krtaylor> #topic OpenStack Program items
18:06:32 <krtaylor> quick mention for the third-party CI documentation WG forming, that will get going next week
18:06:38 <krtaylor> #link https://etherpad.openstack.org/p/third-party-ci-documentation
18:07:06 <krtaylor> and also the monitoring dashboard spec:
18:07:09 <krtaylor> #link https://review.openstack.org/#/c/135170
18:07:47 <krtaylor> and finally meeting times - I'll close out the vote today and email the results
18:07:56 <krtaylor> mentoring times will not change
18:08:16 <krtaylor> work group time will move to  Wednesday starting next week
18:08:40 <ja_> which list are you emailing to?  iirc last time there was a hanging question wrt -dev vs not
18:08:50 <krtaylor> looks like 1500 UTC and 0400 UTC is going to win
18:09:13 <krtaylor> I'll prob hit both, and be sorry for the spam
18:09:44 <krtaylor> all need to know, and some that may not monitor infra alone
18:10:26 <krtaylor> ok, so we can move on if no comments
18:11:29 <krtaylor> skip to open discussion
18:11:44 <krtaylor> #topic Open Discussion
18:11:58 <krtaylor> ja_, any thing you want to discuss>
18:11:59 <krtaylor> ?
18:12:06 <ja_> question - fup/clarification on dougwig's answer last time
18:12:07 <krtaylor> or anyone else?
18:12:34 <krtaylor> ja_, on sizing resources?
18:12:36 <dougwig> ja_: shoot
18:12:47 <krtaylor> hi dougwig
18:12:52 <dougwig> krtaylor: hiya. :)
18:12:59 <ja_> he gave some per-slave stats... having seen "a recently constructed overview diagram", I wasn't sure exactly what slave maps to... compute node?
18:13:43 <dougwig> in the devstack-gate world, the slave runs devstack, so it's all nodes. and usually your hardware runs in parallel to the slave, or if it's a soft appliance, in parallel, or within the same nova.
18:13:57 <ja_> b/c there's a whole pile of table-stakes machines for jenkins et al that I have to size as well.
18:14:09 <ja_> ok I hadn't realized you were talking about devstack
18:14:13 <krtaylor> although the term get overused
18:14:30 <ja_> yeah ... like "resource" in REST land
18:15:17 <dougwig> i have a minimum of three boxes: jenkins (t1.micro at AWS), slave (m3.medium) and our appliance (m1.medium).  and the latter two get added as sets as we need more throughput.
18:15:25 <ja_> does anyone here then have a feel for the size and shape of all the "infra" pieces necessary to drive the compute nodes?
18:15:43 <krtaylor> ja_, in your case, you are looking to figure out how many guests you'll need to run?
18:15:56 <dougwig> i run them all on the jenkins server, which can be quite small.
18:16:16 <dougwig> if youre referring to zuul, gearman, etc.
18:16:17 <krtaylor> yes, the infra services don't require a lot of horsepower
18:17:06 <ja_> well, if I have to do a capital request, I probably get to do about one per year (two maybe in 15 since I'm new at it), so I want to be sure what the fixed-cost piece is too
18:17:18 <ja_> "very small" is a nice answer
18:17:55 <ja_> I had been assuming they'd all be I/O bound hence smallish, but this confirmation helps.
18:18:00 <dougwig> if you want to do a single server, i'd get 4 or 8 cores, 32gb of ram, run jenkins et all natively, and then use vagrant to spawn devstack-gate slaves inside the same box.
18:18:20 <dougwig> et al. i can't believe auto-correct thought 'et all' was more correct.
18:18:37 <krtaylor> ja_, we are driving our system with infra services on guests, but you could probably get by with a nicely equipped PC
18:18:42 <ja_> auto-corr apparently doesn't know latin
18:18:43 <krtaylor> as dougwig said
18:19:39 <krtaylor> we have intel nodes for that (and other uses) next to our PowerKVM nodes
18:20:17 <ja_> I don't expect we'd be running zuul and friends on the same box as the devstack stuff, for the same reason I think powerkvm would separate them...
18:20:41 <dougwig> devstack goes on the slave, which is sacrificial given the security concerns of running mystery code.
18:20:47 <ja_> ...i.e. all the existing doc and setup and experience for the infra bits is x86-oriented, so re-use that community knowledge.
18:20:50 <krtaylor> ja_, not because we couldn't, but it is just easier to be more like upstream services
18:21:10 <krtaylor> exactly
18:21:32 <ja_> true I guess you could run ubuntu now ... I don't think that will be a (supported) option for us
18:21:32 <dougwig> you also want this thing in your DMZ, not inside your cozy little secure network.  and plan on a decent amount (few hundred GB) of slow storage for the logs/web server.
18:22:32 <ja_> right, we have entire first-dot separate network for things that need to be at all accessible from "outside"
18:22:45 <krtaylor> ja_, you can use softlayer swift service for logs too
18:23:15 <krtaylor> we planned a TB for log storage
18:23:22 <ja_> I have been assuming we might - basically on that we should be able to go low-cost.
18:24:07 <dougwig> i think i have a 50GB volume, retain 60 days, currently using 16GB.
18:24:11 <dougwig> for logs
18:24:27 <ja_> iirc the ci page said 6 mo retention
18:24:42 <krtaylor> it used to be 6 months
18:24:47 <dougwig> i thought it was 30 days?  looking...
18:24:48 <krtaylor> they dropped it to a month
18:24:58 <krtaylor> it was taking up way too much space
18:25:10 <krtaylor> that happened in Atlanta
18:25:31 <qwebirc17367> what else happened?
18:25:32 <ja_> hmm, quite sure I read it live on wiki as 6 mo within last 4 wks
18:26:01 <krtaylor> ja_, then that needs to be changed, good patch opportunity
18:26:22 <krtaylor> I'll check it when I get elbow deep in docs next week
18:26:25 <ja_> I know I saved the url for the list.  well first I have to jump the ibm hurdles and get blessed.
18:26:52 <krtaylor> ja_, no easy task
18:27:49 <ja_> is there any historical data on the volume of patches?  I think last time doubwig said it was 30-40ish for nova-only per day, which is quite a bit less than what I'd heard before.
18:28:19 <dougwig> neutron-only
18:28:32 <krtaylor> I pulled together data on that last month for the "core" projects
18:28:36 <ja_> krtaylor, I'm familiar with stds work, have done oss before, so meh just bits on paper
18:28:42 <krtaylor> don't have it at my fingertips atm
18:29:00 <dougwig> and it's about 100-200/day during the busy times, since I need 3 sets of slaves, currently at 18 minute build times.
18:29:37 <ja_> np we you find it, I'll use it.  I'm still flogging pure devstack on my new (VM) sandbox
18:30:09 <krtaylor> ok, anything else?
18:30:18 <krtaylor> if not we'll call it early
18:30:28 <krtaylor> thanks for the assistance dougwig, hope we didn't pull you out of vacation :)
18:31:03 <dougwig> np, i heard my phone buzz, popped over.  :)
18:31:21 <krtaylor> ok, if nothing else then
18:31:47 <krtaylor> thanks everyone, until next week
18:32:02 <ja_> ...that being Wed
18:32:06 <krtaylor> #endmeeting