This repository has been archived by the owner on Jul 16, 2020. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 51
Weekly Meeting 2016 12 01
Tim Pepper edited this page Jan 19, 2017
·
4 revisions
- roll call (1 minute)
- opens (5 minutes)
- bugs (10 minutes)
- triage
- scrub
(09:06:50 AM) ciaomtgbot: kristenc: Error: A meeting topic is required, e.g., '#startmeeting Infrastructure'
(09:06:55 AM) kristenc: #startmeeting weekly meeting
(09:06:55 AM) ciaomtgbot: Meeting started Thu Dec 1 17:06:55 2016 UTC. The chair is kristenc. Information about MeetBot at http://wiki.debian.org/MeetBot.
(09:06:55 AM) ciaomtgbot: Useful Commands: #action #agreed #help #info #idea #link #topic.
(09:06:55 AM) ciaomtgbot: The meeting name has been set to 'weekly_meeting'
(09:07:02 AM) kristenc: #topic role call
(09:07:07 AM) mrkz: o/
(09:07:14 AM) jvillalo: 0/
(09:07:16 AM) rbradford: \o
(09:07:17 AM) kristenc: o/
(09:07:21 AM) tcpepper: \0
(09:07:21 AM) markusry: o/
(09:07:46 AM) btwarden: o/
(09:07:47 AM) _erick0zcr: o/
(09:07:55 AM) albertom: \o/
(09:07:56 AM) kristenc: why you should never think "I have 2 minutes, I'll just check something out really quick on the web"
(09:09:11 AM) kristenc: I have a hard stop time today of :55 past the hour.
(09:09:18 AM) kristenc: #topic Opens
(09:09:38 AM) obedmr: o/
(09:10:13 AM) obedmr: no opens from my side
(09:10:23 AM) mrkz: I guess tcpepper golang1.8 travis topic can be discussed as open?
(09:10:53 AM) kristenc: #topic go 1.8 testing
(09:11:19 AM) kristenc: seems like a good idea to start looking at as long it isn't a blocker, like tim suggested.
(09:11:59 AM) mrkz: yes, would give us some time to fix if something breaks with 1.8 I guess
(09:12:29 AM) kristenc: if there are no objections, should tcpepper change our travis yaml to make it work?
(09:12:38 AM) markusry: tip should catch most of the errors, so it's unlikely we'll find anything, but it can't hurt
(09:12:57 AM) markusry: The only issue is that it might increase the change of random build failures
(09:13:03 AM) markusry: while we still have flaky unit tests
(09:13:10 AM) markusry: sorry increase the chance
(09:13:11 AM) kristenc: ah - true.
(09:13:20 AM) kristenc: now 3 chances to fail.
(09:13:30 AM) kristenc: probability increases.
(09:13:53 AM) ***tcpepper forgets...can we mark the case as a non-blocker? or we humans choose to ignore it and merge things that fail there?
(09:14:19 AM) rbradford: tip failing is already a blocker
(09:14:53 AM) tcpepper: does tip include 1.8?
(09:15:04 AM) kristenc: ok, my inclination is to not do it right now if it adds little value and just increases the odds that our flaky tests will fail.
(09:15:09 AM) markusry: Well tip has 1.8 and newer stuff too I guess
(09:15:19 AM) tcpepper: ah ok then
(09:15:56 AM) rbradford: we should add 1.8 but make tip not a blocker
(09:16:03 AM) markusry: But we'll need to add 1.8 sometime anyway
(09:16:23 AM) markusry: When it's released, assuming we continue to support 1.7 as well
(09:16:23 AM) rbradford: esp if they're starting to merge 1.9 stuff in
(09:17:52 AM) tcpepper: I have one other open
(09:17:55 AM) markusry: Let's add it. We'll need to add 1.8 in the next few months anyway
(09:18:05 AM) markusry: And rbradford is fixing all the flaky tests
(09:18:12 AM) kristenc: right.
(09:18:19 AM) tcpepper: markusry: is there a way to programmatically mark it as non-blocking?
(09:18:30 AM) markusry: In travis?
(09:18:33 AM) tcpepper: ie: ask travis to attempt running it but ignore failre
(09:18:38 AM) markusry: Yes
(09:18:50 AM) tcpepper: I was trolling the travis.yml git log but not seeing it
(09:19:07 AM) markusry: But we wouldn't know there was an issue without looking at the logs
(09:19:44 AM) kristenc: I sort of like having it block.
(09:20:04 AM) tcpepper: ok so just plain add it, fix things as needed, optionally merging some things that break in 1.8 in the near term depending on the specific issue?
(09:20:08 AM) kristenc: we've yet to encounter an issue really where tip was so severely wrong we couldn't make progress.
(09:20:27 AM) markusry: kristenc: I've only seen this once, when golint wasn't working on tip
(09:20:39 AM) kristenc: yeah - let's just see if we wind up having problems and deal with them on a case by case basis.
(09:21:13 AM) kristenc: markusry, and that got addressed fairly quickly, right?
(09:21:28 AM) kristenc: well - at least we worked around it.
(09:21:33 AM) markusry: actually, I don't know. I disabled golint on tip and it's still disabled
(09:21:43 AM) markusry: I should really re-enable it
(09:21:47 AM) kristenc: yes - I was just about to add that that workaround was still in place.
(09:21:59 AM) markusry: I'll enter a bug.
(09:22:09 AM) markusry: There's something else I want to fix in travis as well
(09:22:13 AM) kristenc: ok. so are well all agreed then?
(09:22:38 AM) tcpepper: markusry: ifyou're going to be in there, I'll let you add 1.8 instead of me doing it as kristenc suggested originally
(09:22:52 AM) markusry: tcpepper: Okay, will do
(09:22:55 AM) kristenc: ok.
(09:23:10 AM) kristenc: #action markusry to add go 1.8 checking to travis
(09:23:18 AM) kristenc: next topic?
(09:23:34 AM) tcpepper: I have one other open
(09:24:04 AM) tcpepper: have ya'll seen the failure here: https://travis-ci.org/tpepper/ciao/jobs/180459363 came up in my daily rebasing and is a new one to me.
(09:25:18 AM) albertom: oh so its not only me :P
(09:25:24 AM) kristenc: tcpepper, no, I've not. Is it intermittant?
(09:25:31 AM) ***albertom has the same issue
(09:25:32 AM) tcpepper: so far 2 for 2
(09:25:41 AM) kristenc: did it just start today?
(09:25:45 AM) tcpepper: yes
(09:25:46 AM) albertom: https://travis-ci.org/albertomurillo/ciao/jobs/180456773
(09:26:00 AM) kristenc: ok - something changed in the container no doubt.
(09:26:25 AM) kristenc: dns.start failed d.Start failed d.launch failed exec: \"dnsmasq\": executable file not found in $PATH
(09:26:33 AM) kristenc: implies that dnsmasq is not present.
(09:26:48 AM) tcpepper: yep
(09:27:12 AM) tcpepper: which implies it's now removed from the base image travis uses?
(09:27:19 AM) kristenc: maybe do a test on a fork to see if explicitly getting dnsmasq fixes the problem.
(09:27:29 AM) kristenc: I would guess that to be the case.
(09:28:05 AM) mcastelino [~mcastelin@134.134.137.71] entered the room.
(09:28:12 AM) tcpepper: is there a travis native way to add something? or do you just apt-get install it?
(09:28:28 AM) mcastelino: tcpepper, to install software?
(09:28:35 AM) tcpepper: yes
(09:28:41 AM) kristenc: mcastelino, it seems that dnsmasq has disappeared.
(09:28:54 AM) mcastelino: huh?... wow
(09:28:59 AM) rbradford: weird, yes
(09:29:11 AM) kristenc: mcastelino, https://travis-ci.org/tpepper/ciao/jobs/180459363
(09:29:16 AM) rbradford: my builds have (in the last few minutes) started to fail
(09:29:19 AM) rbradford: for the same reason
(09:30:00 AM) tcpepper: ok their docu says to just before_install: - sudo apt-get install pkg
(09:30:04 AM) tcpepper: i'll pr a quick change
(09:30:22 AM) kristenc: ok.
(09:30:40 AM) mcastelino: did they switch us to the container based image?
(09:32:22 AM) kristenc: no idea.
(09:32:37 AM) kristenc: next open? or shall we do bug triage?
(09:32:48 AM) rbradford: i have an open
(09:32:49 AM) jvillalo: I have a questoin
(09:33:05 AM) kristenc: ok, rbradford wins - what is your open.
(09:33:21 AM) markusry: The version of Ubuntu used seems to have changed since this morning
(09:33:46 AM) markusry: We've gone from Ubuntu 14.04.4 to LTS Ubuntu 14.04.5 LTS
(09:34:13 AM) rbradford: currently we want the owners of files to approve a PR that touches their files. i've got a bot under development that will automatically mention the users on the PR so they know that their feedback is expected
(09:34:25 AM) tcpepper: nice
(09:34:29 AM) rbradford: and it will allow the gatekeeper to know if they're waiting for anyone.
(09:34:37 AM) kristenc: rbradford, cool! sounds useful.
(09:34:38 AM) markusry: Actually, on that topic, I think we should remove sameo as owner of components for the time being
(09:34:46 AM) kristenc: markusry, yes - a good idea.
(09:34:48 AM) markusry: I already mentioned this to sameo
(09:34:55 AM) mcastelino: rbradford, so file based auto add of mention?
(09:34:58 AM) ***sameo feels sad...
(09:35:09 AM) kristenc: sameo, believe me, we are sad too.
(09:35:14 AM) rbradford: mcastelino, yes, i'll move the wiki owners page to a file in git
(09:35:27 AM) albertom: hey thats a good idea!
(09:35:41 AM) mcastelino: that is nice... today unless there is a explicit @xyz we do not know
(09:36:11 AM) markusry: sameo: You could become the owner of qemu
(09:36:27 AM) markusry: I mean ciao's qemu package
(09:36:36 AM) sameo: markusry: Sounds good to me.
(09:36:39 AM) albertom: ansible repo has a bot that does cool stuff too like auto tagging the issues and pinging the owner every once in a while so he doesnt foget the issue
(09:36:40 AM) tcpepper: I'm willing to take ciao-cli and payloads. I nominate wdouglas for ciao-cert.
(09:37:09 AM) markusry: any changes we make to this will affect cc so you'll need to review them anyway
(09:38:09 AM) kristenc: #topic maintainer bot
(09:38:46 AM) kristenc: #info rbradford has written a bot to mention maintainers who need to review PRs
(09:39:05 AM) tcpepper: fyi: added dnsmasq still has trouble https://travis-ci.org/tpepper/ciao/jobs/180466435
(09:39:17 AM) kristenc: #action rbradford to add maintainers file to git
(09:39:29 AM) kristenc: #action sameo to be removed as owner of everything but qemu
(09:39:37 AM) kristenc: anything I missed?
(09:39:41 AM) sameo: kristenc: Who's taking over ssntp ?
(09:39:55 AM) kristenc: tbd?
(09:40:05 AM) kristenc: we don't have to find owners for everything.
(09:40:20 AM) kristenc: the gatekeeper will be reviewing in any case.
(09:40:33 AM) kristenc: why not wait till it becomes obvious who's been poking around there.
(09:40:37 AM) piet [~textual@134.134.139.76] entered the room.
(09:40:40 AM) pixelgeek [~cnorman@192.55.54.40] entered the room.
(09:40:51 AM) mcastelino: SSNTP has not changed in a while.. so maybe sameo can still ACK/NACK :)
(09:40:53 AM) wdouglas: Hrms
(09:41:28 AM) markusry: kristenc: The merging guidelines say that a review from an owner is needed to merge a patch
(09:41:43 AM) kristenc: markusry, presumably, that's only if there is an owner.
(09:41:54 AM) kristenc: in the event of no owner, the gatekeeper reviews.
(09:42:05 AM) kristenc: (which they do anyway)
(09:42:07 AM) markusry: Okay, sounds okay to me
(09:42:10 AM) sameo: mcastelino: I'm ok with that.
(09:42:48 AM) kristenc: things that don't change much or have trivial changes probably don't need dedicated owners.
(09:44:15 AM) kristenc: ok, jvillalo you had a question.
(09:44:20 AM) jvillalo: right
(09:44:26 AM) jvillalo: Regarding the external api impl. This one is only available through the new api mechanism right?
(09:44:26 AM) jvillalo: So just to confirm the workflow, we'll need to 1. get resources, 2. find external api within the list of resources that webui needs to handle. 3. make requests to api
(09:44:50 AM) jvillalo: find the external IP endpoint I mean (step 2)
(09:44:57 AM) kristenc: jvillalo, the external IP api?
(09:45:09 AM) jvillalo: endpoint list
(09:45:26 AM) kristenc: jvillalo, I tried to implement this the way I thought might be good in ciao-cli
(09:45:32 AM) jvillalo: so the ones for listing ips, handling pools, etc..
(09:45:36 AM) kristenc: I tried to utilize the discovery mechanism
(09:45:39 AM) kristenc: (the links)
(09:45:44 AM) kristenc: to not hardcode the paths.
(09:46:13 AM) kristenc: so - if you take a look at that code, you'll see that I start at / or /{tenant_id} for not privileged user.
(09:46:18 AM) kristenc: it lists the resources you have.
(09:46:30 AM) kristenc: I look for either pools or external-ips, and use the href to go from there.
(09:46:45 AM) kristenc: does that answer your question?
(09:46:55 AM) jvillalo: excellent, I'll check both the / and /{tenant)id} nad cli code as well
(09:47:03 AM) jvillalo: yes it does, thanks a lot :)
(09:47:22 AM) kristenc: jvillalo, for the versioning, you have a choice of asking for an explicit version (via content-type)
(09:47:30 AM) kristenc: or asking for just plain old json.
(09:47:41 AM) kristenc: if you ask for json, you'll always get the oldest version of the resource.
(09:47:55 AM) kristenc: which at the moment is the same as the newest version :)
(09:48:24 AM) jvillalo: sounds good, thanks
(09:48:52 AM) kristenc: jvillalo, albertom had a good question about the keystone catalog and the new ciao http api port.
(09:49:35 AM) kristenc: should we add a catalog entry there to help discover the service? or would you even use it? it's not an openstack service, so this would not be for compatibility at all, just if you think it would be a good idea for consumers.
(09:49:35 AM) albertom: do i ?
(09:49:42 AM) jvillalo: oh, sorry I wasn't aware of that question, pelase shoot
(09:49:57 AM) kristenc: jvillalo, do you use the keystone catalog?
(09:50:01 AM) jvillalo: yes
(09:50:03 AM) jvillalo: I do
(09:50:12 AM) kristenc: jvillalo, should we add the new ciao http api to it?
(09:50:23 AM) kristenc: and what should we call it :).
(09:51:07 AM) jvillalo: Let me see if I understand correctly, so ciao controller's catalog would list keystone endpoints(including keystone's catalog)?
(09:51:13 AM) albertom: are we using openstack api for floating ips? or have we gone in other direction by now?
(09:51:15 AM) jvillalo: is that correct?
(09:51:21 AM) kristenc: jvillalo, no.
(09:51:42 AM) kristenc: jvillalo, you know how you can go ask keystone where all the ports are for various services (image, cinder, nova)?
(09:52:21 AM) kristenc: we were wondering if we should similarly add one for the new ciao http api and call it some service name.
(09:52:27 AM) mcastelino: albertom, we are not using the openstack API's or the openstack flow for floating_ips... the verify.sh script will give you a good idea of the flow as it stands today
(09:52:48 AM) kristenc: we are not doing "floating ips" :).
(09:53:14 AM) albertom: oh external ips right :P
(09:53:18 AM) albertom: i'll check
(09:53:26 AM) jvillalo: Oh I see, sorry I think I miss-understood the question, so here is what happens, what we do ask keystone is the list of the tenants/projects only, and that helps me know how to call other endpoints because it's the only way I know the tenant/projects ids
(09:53:32 AM) jvillalo: that's the only we heavily rely on
(09:53:37 AM) jvillalo: today.
(09:53:44 AM) kristenc: jvillalo, ok - so you don't use it for service discovery.
(09:54:06 AM) jvillalo: Not right now, no.
(09:54:13 AM) kristenc: well - I'm not adding it if no one will use it. we can add it later if we decide it's a good idea.
(09:54:31 AM) jvillalo: Now, my opinion is
(09:54:49 AM) jvillalo: we won't need it until's services number grow higher
(09:55:20 AM) jvillalo: so we started only with keystone and compute api at the begining on the year, effort to implement discovery of only 2 services wan't worth it yet
(09:55:21 AM) kristenc: how do you know which port to go to if it's configurable?
(09:55:55 AM) tcpepper: kristenc: don't forget your hard stop
(09:56:13 AM) jvillalo: We don't discover it today. So we don't we rely on default ports
(09:56:22 AM) kristenc: yes, thank you. well sorry, unfortunately I need to leave now. we did zero bug triage, let's try to get that out the way another time.
(09:56:23 AM) jvillalo: unless we know it's production environment
(09:56:29 AM) kristenc: #endmeeting
(09:56:29 AM) ciaomtgbot: Meeting ended Thu Dec 1 17:56:28 2016 UTC. Information about MeetBot at http://wiki.debian.org/MeetBot . (v 0.1.4)
(09:56:29 AM) ciaomtgbot: Minutes: ciao-project/2016/ciao-project.2016-12-01-17.06.html
(09:56:29 AM) ciaomtgbot: Minutes (text): ciao-project/2016/ciao-project.2016-12-01-17.06.txt
(09:56:29 AM) ciaomtgbot: Log: ciao-project/2016/ciao-project.2016-12-01-17.06.log.html
Development
Architecture