#metabrainz

potential metadata customer maybe? someone should test de waters

~the

Freso: Yeah, I am going, but it would be cool to see the MusicBrainz folk at a conference sometime. :) Maybe what might be more realistic is going to FOSDEM next year, since I think most of you aren't that far from Brussels.

zas: ping

I agree that the script exiting is the problem at hand. But exec is the right thing to do since it replaces the current process with the new process that gets loaded. in this case, python. it works fine for all the other services that run out of that one image.

at least the containers is back to restarting, which is an improvement. :)

thanks for the magic incantation to fix that

now to figure out why it dies silently. :(

iliekcomputers: any ideas?

Mineo: ping

pong

I just read your comment on PR 657

the fixes are somewhat independent, but given the time it usually takes for one PR to get merged, I preferred to join all "cover-art-related" fixes in one PR and not having to fix conflicts and everything

ruaok: pooong (in which timezone are you ?? ahahah)

also, the commits that fix each one are independent

finally... I've been trying to reproduce your issue, and still couldn't

antlarr: having too much code in one PR usually don't help for fast merges ;)

zas: hehe

that said, i would like to merge those fixes for 1.4.1 so i can live with this (not so) fat PR

Mineo: i think about merging the config option PR too, since it doesn't include any translatable strings and shouldn't impact anything, what do you think ?

Mineo: about your remaining issue with PICARD-1001, do you see the difference in the coverartbox? or in the infodialog?

PICARD-1001: Cover art changes are indicated even if there's no change after loading a release https://tickets.metabrainz.org/browse/PICARD-1001

Zas: further than iliekcomputers... :-)

Mineo: all I could reproduce with your configuration was http://i.imgur.com/gn27Agu.png, but if you see, the cover art box is ok (I'm having a look at the info dialog now)

neither shows a difference

ruaok: did you find out why it's exiting with no logs?

antlarr: also, I think you need to save it with that config at least once - there shouldn't be a back image with it

No, busy drinking 🍺 atm

Mineo: right, now I don't have any difference

ha

so, what's the remaining problem with PICARD-1001 ?

PICARD-1001: Cover art changes are indicated even if there's no change after loading a release https://tickets.metabrainz.org/browse/PICARD-1001

and you still have the green checkmarks next to each track after removing the release and loading it again?

ah, I see

ruaok, iliekcomputers: did you test anything yet to get the error reported ? i think just adding "exec 2>&1" to your init script might help

there's a green box

it should be a green checkmark

right?

yes

I was looking at the coverartbox/info dialog the whole time...

also don't forget "docker events" command may help to see what is going on with fast dying containers (it doesn't help much in this case)

ok, I have like 15 minutes this morning to work on this... let's see if I find it quickly

Zas: there are 3 other services that give proper output using the same exec statement.

I really doubt it is the exec

ruaok: but the other services don't fail, right ?

2017-03-19T10:35:09.810906857Z container die 1d454282ac1c6b1ae7e8d084d2a4f528cedfe774e3d43a31bff39d25e28cd81b (exitCode=0, image=metabrainz/listenbrainz, name=listenbrainz-influx-writer)

exit code 0 ??

so it means the script just exits without error

zas: re config file PR: I'll never say no to getting one of my PRs merged ;-)

Correct, fail without error.

if the python script does that, it is expected the container dies, it shouldn't exit but on signals (or serious failing)

There are only two places from where the script can stop

if it fails to write after x retries

and here: https://github.com/metabrainz/listenbrainz-serv...

iliekcomputers: be sure to output a message when the script exits (or start)

zas: both places have messages :/

but not to stdout, right ?

no

and re PR 657: I haven't looked at the code yet, but the amount of hasattr and getattr calls in it is already scaring me :(

Mineo: ahah ;)

iliekcomputers: the problem is that if it logs to a file in a container that only exists during the time the command runs... it is pretty pointless. So ensure you can actually see when the script starts and ends in the docker logs command

you can get the container id from docker events for fast-dying containers and docker logs <id>

also test something is visible in logs in case of syntax error in the script too

iliekcomputers: let's add some stupid simple stdout logging and see if that improves?

also, handle signals in your script for clean stop

btw, the lack of proper handling of container stop conditions is a real pain for me, since, for example, if i need to reboot system or upgrade docker command, bad things may happen if underlying stuff isn't managing container shutdown properly.

Mineo: the file is marked as "with pending changes" because file.orig_metadata.images is empty (since the file doesn't have any embeeded image because your options enable "save cover images as separate file"), but file.metadata.images have a cover image, since it's downloaded from the release

so in file.update(), the state is set to CHANGED

I'm afraid I don't have time to fix it today (maybe tonight, but I'm not sure), so if you want, tell me how you'd fix (loading file.orig_metadata.images with images from the local files provider when there are no embeeded images?) and I'll try to fix it tomorrow

[listenbrainz-server] paramsingh opened pull request #150: Add logging to stdout to influx-writer (master...influx-writer/logging) https://git.io/vydGB

Zas: We really need to create a docker best practices doc for meb.

I didn't realize that this is what you meant by signals earlier.

iliekcomputers: thanks for the pr. Will look when I get onto the metro.

This city is crazy. :-)

ruaok: no prob, happy travels :)

:-)

Ahh, perfect. Will read when sober. :-)

ruaok: in 2019 then ? ;))

antlarr: I'm afraid I don't know how any of that code works anymore

Ok ok ok. Sober-ish. Well, more sober than now.

And, why do you think I would be sober in 2019???

ruaok: just joking ;) i hope you'll be sober before that ;)

zas: phhhhhhhhffft.

anyone about who can help with 'sir'? followed the docs to try and get it set up so I can try mbsssss-based search

but it looks like they're out of date

as I get "ImportError: cannot import name init_raven_client"