mozilla :: #conduit

9 Aug 2017
13:44ckolosmorning guys, still working on the state of the repo in dev
13:55* imadueme
14:11imaduemedkl: I've updated Also the warning you mentioned yesterday seems benign
14:12davidwalshckolos: ping re: csp endpoint?
14:20dkl|afkimadueme: ok
14:43zalundo we have a vidyo standup today?
14:45imaduemeI'm cool with either vidyo or irc.
14:49zalunlet's make it IRC then
14:49davidwalshMy Vidyo taking eternity to come up
14:50imaduemeMy standup is that I will have a patch for lando api by lunch time - zalun I will probably set you as the reviewer on that. Then I will be using that work to finish the UI. davidwalsh, dkl: Any way we can help the phab/bugzilla work without writing php/perl :) ?
14:51zalunI'm working on fake-transplant and pingback in lando-api. I wait for ckolos to make working again. When I'll finish this I will work on phabricator to have Bugzilla URL specified in environmant variable or find out about global config variable.
14:52zalunimadueme: please ping me about it - working weird hours today
14:52dklstandup: landed change yesterday to stop building docs on every mozphab container start and instead add argument to docker-compose run that does it on demand. Working on finishing reviewing mars/imadueme changes to demo for showing off lando. Also going to work on BMO code for setting flags when revisions are accepted/rejected.
15:07davidwalshimadueme: PR in for CSP; mars: updated my flaskpyoidc PR
15:07davidwalshSavage mars, savage
15:11imadueme+1 davidwalsh, will leave for a while if ckolos or another person want to take a look.
15:11imaduemeleave the pr that is
15:15imaduemedavidwalsh: I know >:D
15:16imaduemeIt is all part of my plan to get more points that anyone else on the trello board. mwahahahahahAHAHAHAHAHA!!!!!!!!!!
15:28davidwalshMy son's class has a field trip to the local splash pad which is a few minutes away; going to make a quick cameo; back in a few
15:28ckolosNew lando revisions deployed to dev
15:29imaduemeThanks, ckolos!
15:29ckolosimadueme: Might be useful to set a content type on this endpoint:
15:29ckolosapi sets it, but ui doesn't.
15:31imaduemeckolos: Inspecting the network response shows the content type header for me. (It is weird because both the API and UI implement that endpoint with exactly the same code)
15:32ckolosre:, the mozilla-central repo on this instance is in such a bad state that it's basically useless. All attempts to manually update the repo are failing with a helpful "transaction abort!" message which then leaves the repo in need of 'hg recover'
15:32ckolosimadueme: that is odd
15:32imaduemeNo worries I'll look into it :)
15:35ckolosimadueme: my console shows this interesting mix of colors:
15:37ckolosmore re:, I am working on re-importing the repo via an hg bundle right now and will try to slide that into place; difficulty level: I have a dr appt in 3 hours that I can't resched
15:37ckolosso I will bring the host back up, but be aware that mozilla-central is in a known-bad-state on that host.
15:39imaduemeckolos: Love that color pallet. As far as goes: Do you have any idea what could have caused the problem? Is clustered mode / ebs working against us?
15:39ckoloswe're not clustered
15:39ckolosso that's not it.
15:40ckolosafter talking to gps yesterday, I think that the way I configured the base container was incorrect
15:40ckolosthe container was set to timeout=0
15:40ckoloswhich is essentially a sigkill on anything running in there
15:40ckolosso if a daemon is running to update the repo, it's not given a chance to stop and clean up
15:41ckolosone of the inconsistent complaints from the daemon has been waiting on lock from a non-existent host
15:41ckoloswhich was really messing with me mostly b/c it was so intermittant
15:42ckolosafter noticing that the repo status has remained in a bad state for more than a day, I started digging into the locks on the repo
15:42ckolosand found a couple of stale lock files; digging through cloudtrail, I was able to match those to previous mozphab hosts
15:43ckolosso that side of things is somewhat fixed
15:43ckoloslock files removed, etc.
15:43ckolosbut the repo itself is unable to update itself, even from the command line
15:43ckoloshg runs for a while, but then dies out with "transaction abort!"
15:43ckolosand boom. no explanation, just an abort.
15:44ckolosanyway, it's been fun and I'm really sorry that this is in this state.
15:45ckolosthat's what I see
15:46imaduemeOh man, sorry for the trouble :). Maybe gps or smacleod have an idea. As far as the host, I would say you can just leave it down until the repo is fixed. That's not really my call, but, my 2 cents. We always have the demo environment if we want to use a Phab instance for development.
15:48ckolosimadueme: gps was very helpful to me yesterday
15:48ckolosOn the plus side, we now deploy the latest mercurial with the image, so that's a good thing
15:50smacleodckolos: how would I go about sshing into the box?
15:52ckolossmacleod: ssh
15:52ckolosthen ssh
15:52ckolosneed to add you
15:55ckolossmacleod: after the bastion, ssh to
15:55smacleodckolos: no hostname?
15:55ckolosI suppose you could use the internal one
15:55smacleodnot much different than the IP itself heh
15:56ckolosright ;)
15:56ckolosthe busted repo is in /phabricator_storage/1
15:56ckolossmacleod: you should have full sudo
15:56smacleodckolos: awesome, I'm in, thanks
15:57ckolossmacleod: that's on the "localhost"
15:57ckolosnot in the container
15:58smacleodAh, right, containers haha
17:19davidwalsh|brbimadueme: mars: Since you've both touched lando-ui more than I, wanna check out the "Coding rules" section here?
17:23imaduemedavidwalsh: We're doing most of those recommendation, but, not all. I think it is something we'd have to discuss as we get closer to release
17:43ckolosall: smacleod has graciously volunteered to look at the repo issue; I have to run out for an appt, but I've asked smac if he can restart the required things once he's done. I will have my phone, so ping if you have any questions
17:48imaduemeckolos: Thanks, hope the doctor has good news for ya :)
17:48davidwalshzalun: ping me when you'r earound
17:49zalunimadueme: ping
17:50imaduemezalun: Was that for davidwalsh? :D
17:50zalundavidwalsh: ping
18:13mcotedkl: thinking more about the security stuff... something I don't think we considered is a person posting a revision to a bug they don't have access to
18:14dklmcote: yeah davidwalsh were thinking about that. i would think we would kick it back
18:14dklrequire that the person can see the bug
18:14dklin BMO we cannot attach to a bug we cannot see
18:15mcoteyeah exactly
18:15dklok part of the plan then
18:15mcoteI guess technically there's nothing wrong with posting a revision like that, assuming they don't get access to the bug, but I think that would be confusing
18:15mcoteso yeah, requiring the access to the bug makes sense and is simpler
18:15mcoteand clearer
18:16davidwalshmcote: I was about to take ; was talking to dkl about that via DM
18:16davidwalshmcote: But also with dkl out, maybe I should focus on the perl stuff
18:16davidwalshJust trying to figure out with dkl the best plan
18:16davidwalshI don't know what others are working on
18:16mcoteyeah I'm not sure tbqh
18:16mcoteanything related to BMO-Phab integration
18:16mcoteif you are doing perl, you could get dylan to review while dkl is away
18:17mcoteor glob for that matter
18:17mcoteI think the one you mentioned is better though
18:18mcotebest to finish the bits related to security
18:22dylandefinitely, I'm just in a minor hell signals right now
18:24kthiessenmcote: Trying to pull up results in an HTTP 502 for me -- do I need to be on a VPN or somesuch?
18:24mcotekthiessen: apparently there are issues with our dev cluster today
18:24mcotefor the first time since it was launched :\
18:24kthiessenmcote: So my timing remains tremendous, then. Good to know.
18:24mcoteheh yup
18:24mcotewe're working on restoring it
18:24kthiessenmcote: I tell you, it's the fact that the QA person is looking.
18:25mcotehaha at least now we know the root cause
18:25kthiessenNothing brings out the corner cases faster.
18:25davidwalshI'd forgotten the joys of github source searches within phabricator
18:25kthiessenLet me know if there's anything I can do to help, please?
18:27kthiessen(Was able to log into bugzilla-dev with Github auth no problem, btw.)
18:35mcotekthiessen: smacleod & ckolos are on it; I'll let them figure out if they need any help :)
18:36smacleodno help needed at this point
18:36smacleodfixing things is slow going, commands are taking forever to complete
20:14davidwalshWell shish...I'm going to have to figure out how to get the login provider config values to get the bugzilla domain and protocol
20:40davidwalshdkl: Did you modify any PHP to use the BMO user or API key as a header for requests to BMO?
20:40davidwalshHaven't you done something like that?
20:44dkli do that as part of a build step in the Update Revision test plan.
20:44dklbut it is a hack in that it passes the phab-bot api key in the HTTP basic auth header
20:45dklso it didnt require php code changes and just use the functionality already there
20:57davidwalshdkl: Where/how in phabricator do we store the phab-bot info? I'll need its API key (I'm guessing) to programmatically check that a bug id is valid, etc.
20:58davidwalshdkl: Also, is there someone we can harass to update the phab permissions so I can write to files? mounting a volume makes phab throw 502s at every endpoint fml eventually loaded
21:11dklif you want to review that i go ahead and merge it
21:16dkl|afkdinner time
10 Aug 2017
No messages
Last message: 12 days and 8 hours ago