mozilla :: #conduit

8 Aug 2017
14:44dklall: IRC standup again today? any blockers?
14:46zalunI'm OK with IRC - no real blockers.
14:46imaduemeToday I'm continuing the lando ui commit message preview. I'm going to separate the card into 2, with the new one being the api endpoint that formats the commit message so that that logic will all live in lando api. No blockers for me, I do a first review of stuff anyone has.
14:47imaduemeI can* do
14:49davidwalshI'm catching up on email
14:53imaduemedavidwalsh: About your CSP card for Lando UI, I believe ckolos added a csp report endpoint via the nginx configuration. I wasn't sure if you already knew about that, so wanted to give you a heads up!
14:53dklstandup: finish reviewing mars PR for adding lando to demo.
14:54davidwalshimadueme: that case, the card should be done. Can you confirm ckolos ?
14:54imaduemedavidwalsh: I think we will still need to add the URL in our CSP headers, correct?
14:54ckolosthe reporing endpoint is part of nginx, yes
14:54davidwalshimadueme: Ahh yes, good call, sorry
14:55imaduemeGlad to help :)
14:56imaduemezalun: Do you mind if I take, it seems to be coupled to my new card I just made
14:56zalunimadueme: yes - mcote mentioned it
14:56imaduemezalun: Cool, thanks!
16:19ckoloshey guys, not sure what to make of this error I'm seeing on the dev instance (i.e. is this a phabricator problem or something funny in our hg repo):
16:21smacleodgps: ^
16:22smacleodckolos: I've never seen that, gps might know. If he's not around later I'll look into it
16:22gpsyou got that over https://? impressive
16:22gpsthat usually means a corrupted packet
16:22ckolosI blame php
16:22gpsbut if TLS is in play
16:23gpsi assume you are using mercurial 4.2.2?
16:23gpsif not, please upgrade
16:23gpsthere were some patches in 4.2 to better report network-related failures
16:24ckoloshg --version shows 3.8.2
16:28ckolosIt looks like we might be needing to do some different things in the container guys
16:28ckoloslike starting over with a new base
16:29gpsi can look into getting alpine linux support into mercurial's build system
16:29ckolosit's weird
16:29gpsthey already have make targets for producing rpm and deb packages inside docker
16:29ckolosalpine has a 4.2.2 package available on their website
16:29ckolosbut the container doesn't show it
16:30ckolosah, I see, it's in edge
16:31gpswell, you should be all set then
16:31ckolossort of
16:31ckolosit's a possiblilty, but the base container is alpine by way of the php project
16:31ckolosso I need to track down which version their base image is
16:32gpsmy experience with docker images is that you inevitably fork so you can do nice things
16:32ckolosyeah, their based on 3.4
16:32ckolosthey're even
16:32gpsi wish docker would focus on enabling image creation from recomposable recipes than from base images
16:32gpssomething... like puppet or ansible
16:33gpsor packer
17:09dklimadueme: seen that before?
17:09dkltrying to build the lando-ui image using the same steps from circle.yml
17:11dklnm. something i did to docker/Dockerfile-dev
17:11dklreverting fixed it
17:11imaduemeAh, cool :)
17:12imaduemeLet me know if something else comes up
17:12dklok. so running invoke build gives me an error now
17:16imaduemedkl: Never seen that before either, is that on mars' demo branch or just master lando ui?
17:22dklimadueme: docker-compose build works fine. it is using docker/Dockerfile-dev.
17:23dklbut invoke build is using docker/Dockerfile-prod which is where I get the error
17:23dklthe two are very different
17:23dkldoes Dockerfile-prod build for you locally?
17:24imaduemedkl: I just ran it and it builds for me locally output relevant:
17:26imaduemedkl: Ok, yeah. It works for me, hmm
17:28dklok something environmental then. will poke around a little more
17:29dklimadueme: also for the demo work, I can only get landui to work if I go to port 9000. The PORT=80 in docker-compose.yml doesn't see to be having an effect
17:32imaduemedkl: Is this the prod container or dev? (I assume dev, but, that's weird because the docker-compose for dev defaults to port 7777.)
17:38imaduemeoh dkl: if the disk which you are running the `invoke build` command doesn't allow for symlinks that would make sense. I.e. if your vm's drive is an nfs drive, samba share, or virtualbox shared folder.
17:39dklimadueme: none of those unfortantely
17:39imaduemeWorth a shot :)
17:40imaduemeAnyway, about the port problem. If you change the port, you also have to change the port ending on the SESSION_COOKIE_DOMAIN variables (2 of them) dkl. (If that's not the problem, then, we'll have to dig further)
17:41imaduemeI will make sure to update the lando ui readme with all this
17:50dklwill try that
17:58gpsckolos: regarding docker, watch out for aufs: mercurial is capable of causing filesystem corruption on aufs due to concurrent I/O operations
17:58gpsthis will manifest as "file doesn't exist" errors
17:58gpsaufs is a crap filesystem. avoid it at all costs.
17:58gpssee also bug 1291940
17:58firebot NEW, TaskCluster slows down considerably when using AUFS
18:06ckolosgps: we use raw disk via devmapper and lv
18:09dklimadueme: no go. i removed the :9000 from the SECURE_* vars to match PORT=80 but it couldn't connect
18:09dklhttp://lando-ui.test:9000 works if i do not change anything
18:09dklhere is the new docker-compose.yml
18:12imaduemedkl: Let me try it out
18:13imaduemefwiw, in that PR those 9000's should definitely be removed from the Session keys
18:17imaduemedkl: Oh, it looks like the ui process died
18:25circleci-botFailed: ckolos's build (#53; push) in mozilla-services/mozphab (ckolos/update_mercurial) --
18:26ckolos"no space left on device"?
18:29ckolosmeh, builds without cache
18:31circleci-botFixed: ckolos's build (#54; retry by ckolos) in mozilla-services/mozphab (ckolos/update_mercurial) --
18:32ckolosdkl smacleod: can you r? the PR for that change above?
18:37smacleodckolos: we should maybe use a hashed requirements list since this is building the production image
18:38ckoloshashed requirements list?
18:38ckolossorry, I am teh n00b
18:38dklimadueme: oh yeah. I also had to add VERSION_PATH=/app/version.json to the lando-ui section to keep lando-ui.test from crashing
18:38dklthat was going to be in my review
18:38imaduemedkl: Yup, I'm actually fixing the whole thing now :)
18:39smacleodckolos: e.g.
18:39ckolosah, okay, sure
18:39smacleodSee pip require-hashes
18:40smacleodAlso "hashin"
18:40smacleodckolos: ^
18:50davidwalshdkl: I approved your PR
18:50davidwalshdkl: Any insight as the R+ stuff? Have you started that?
18:55circleci-botSuccess: dklawren's build (#55; push) in mozilla-services/mozphab (master) --
19:00imaduemedkl: I set you as the reviewer on, it should work the first try, fingers crossed.
19:08circleci-botSuccess: ckolos's build (#56; push) in mozilla-services/mozphab (ckolos/update_mercurial) --
19:15circleci-botSuccess: ckolos's build (#57; push) in mozilla-services/mozphab (ckolos/update_mercurial) --
19:26ckolosall: I need a short window on the dev mozphab instance for a image update
19:26ckolosanyone object?
19:28imaduemeNot I
20:25dklckolos: nope
20:28dklimadueme: getting some seemingly harmless warnings 'Contents of /etc/nginx/conf.d/default.conf did not change. Skipping notification 'nginx -s reload''
20:30ckolosI think the local observed repo is in a bad state
20:42circleci-botSuccess: dklawren's build (#58; push) in mozilla-services/mozphab (master) --
21:15ckolosmozphab is down in dev until I can re-verify and recover the repo state; it's basically fubar right now.
21:16ckolosI am trying to not give into my misgivings and quit, move, and change my name.
21:18ckolosHumor is low in this room today I see
21:27dklckolos: oh i get it. LOL
21:27ckolostoo little too late man ;)
21:30ckolosokay, repo is back to a "good" state
21:30ckoloshost is back up
21:31* ckolos crosses fingers
21:34ckolosdkl: r+
21:49davidwalshckolos: Where can I find the nginx config file to get the CSP reporting endpoint?
21:51ckolos_davidwalsh: ping me in the morning and I'll pb them for you
21:51davidwalsh"Take two pb's and ping me in the morning"
21:51davidwalshckolos_: OK, thanks!
9 Aug 2017
No messages
Last message: 13 days and 15 hours ago