#metabrainz

yes makes sense.

lucifer: there is a setting to log disconnections, but it's not enabled

Would you guys mind if I create a smol alternate MusicBrainz docker instance on wolf? If so, how do I do it without interfering with the existing instance?

from gaga in tmux, I run `ssh -L 55432:127.0.0.1:5432 robert@wolf`

Pratha-Fish: the container I made is running with --network host, so it can access any ports exposed on the host

I'm not sure which port the database you want to access is on, though

bitmap: it's on 5432

mayhem: docker container won't be able to access localhost ports i think.

not sure if we run with --network=host on prod.

*5432 until and unless I start another musicbrainz-docker instance different from the existing one

sorry forgot to mention I am not running on docker.

just a venv. ends up being easier for tunneling

ah cool then

Pratha-Fish: then try psql -h localhost -p 5432 -U musicbrainz -d musicbrainz_db

looks good to me

bitmap: yes, I've been using that one for a while now :D

forgot to fetch the right password.

Pratha-Fish: I mean it should work inside the python container too

whoops

query started. monitoring wolf.

Pratha-Fish: also, I mounted your home dir as /snaek inside the container

bitmap: thanks that worked!

thanks a ton 🥹

try `su snaek` in the container and then cd /snaek

I set the uid/gid to the same as the host

mayhem: uhh, i think we also lost the messybrainz table in that issue.

its going to be another PITA to restore that table.

oh joy.

how so?

uuid are generated randomly and not from the source data.

but we backup all those tables with pg-dump no?

bitmap: I can only see directory called "csp-errors" in /snaek

oh that might be my home dir haha. sorry, let me fix

mayhem: i just checked and no. msb table is not backed up.

O_O

wow.

we need to review our backups. this is bad. :(

indeed

SELECT * INTO ?

well not *, but msids

Pratha-Fish: now it should be fine

yeah msids, and the 4-5 fields that go into. plus deduplicating with all the data that has accumulated in the past 4 days

ok, how about this:

1. stop TS wrtier

2. SELECT INTO

bitmap: working!

3. REstart TS writer

4. Write script to fix up 4 days.

yup that's the plan.

ok, can we get started with 1-3 asap>

?

we could but its better to run in the morning. otherwise the ts writer will have to be stopped overnight.

ok, fair.

its late enough as is.

lol

there's one more unbacked up table that is going to be hard to restore. mbid_manual_mapping.

I just ran the partial set of the mb-metadata cache data.

and sure enough data is now being returned, for those 3 artists. that's a good sign.

hard?

where would we get the data from?

it would be impossible. but we have it mixed in spark dumps.

🤯

this is going to be a fun week, isn't it?

compare the automapper assigned with the recording mbid present in the dumps, if its different and not present in the actual listen submitted to LB then it was a manually mapped one.

so have to compare three datasets to recover that.

🤯 🤯 🤯

i am amazed that postgres let that query delete all tables in the script but kills our cache queries before completing execution every time.

might be a level of effort optimization

lucifer: regarding the pgbouncer logs, I can see why we disabled them, since it logs an entry every 0.005 seconds or less (and usually the info is not useful)

LB-1392

LB-1392: Document and require all tables to be dumped https://tickets.metabrainz.org/browse/LB-1392

bitmap: i see, is it possible to change that setting without restart?

I've started on the PR to dump all tables.

yeah, I can make it temporarily log disconnections

but it will probably just say "client close request" without any other info

cool, let's get on that tomorrow.

can we identify the connection that closed down?

I'll start working on it now.

it shows something like musicbrainz_db/musicbrainz@172.17.0.1:45668 closing because: client close request (age=0s)

mayhem: yeah that is one part, other is to error in say CI if we add a new table and forget to dump it.

those docker ports are completely opaque afaik

i see yeah. :/

good idea

thanks for checking that bitmap.

np

you are disabling the statement_timeout and pg is killing it for some other reason? (if pg is killing it, it should be in the pg logs, anyway)

lucifer: do we dump mapping.mb_metadata_cache ?

a lot of generated data to dump.

lucifer: while we're at it, we should switch to zstd compression. that is going to make dumps and moving data to spark a lot faster.

does tmux work with docker? 💀

Pratha-Fish: in what way?

I did the following:

mayhem: I would probably not dump generated data.

But rather make it reliable to generate

Logged into wolf > entered tmux > entered the python 3.11 container > su snaek > started_some_job

ok, then I see that we need to add "messybrainz.submissions" and "mbid_manual_mapping"

Pratha-Fish: yes that's fine

Then I exited tmux by pressing CTRL + B > D

Makes sense

Now, that thew me back to home dir as expected

but when I entered tmux again, I expected to go back inside the docker container and see updates on the job I was running

But entering tmux again just got me to the home dir

tbh I don't remember how to re-attach with tmux (I use screen)

NP, I'll try to figure something out and come back to you if it doesn't work :)

are you sure it's not just opening another session? can you switch sessions?

verifying

session or window I guess

[listenbrainz-server] 14mayhem opened pull request #2620 (03master…LB-1392): LB-1392: Add missing tables to dumps and ensure that all tables are dumped https://github.com/metabrainz/listenbrainz-serv...

Yes, I connected to the same session :D

Still no luck

Looks like I'll have to install tmux inside the container

Hope I didn't kill docker on wolf https://usercontent.irccloud-cdn.com/file/TP0Uk...

no you just stopped the container

phew seems like it

Sorry for the silly question again, but how do I restart it?

docker start snaek-python

thanks 🤦

sorry I'm not sure about your tmux issue, I know the same works fine in screen. if you wanna try installing tmux in the container, go ahead

bitmap: just figured it out :D

as simple as installing tmux on the container

and attaching to the session after entering the container

good :)

:)

bitmap: looks like we have something tangible on our hands right now :)

Not to mention reosarevok

Here's the whole repo: https://github.com/Prathamesh-Ghatole/MusicBrai...

I haven't created any PRs yet because the whole process of figuring out to even what to do in this project was pretty hectic. Will host the rest of the code on https://github.com/MetaBrainz/MusicBrainz-AreaBot/ ASAP.

`tmux a` re-attaches to an existing session btw

Just running tmux starts a new session

atj: thanks, I was making the same mistake lmao. now I just use tmux attach -t sesh_num

E p i c

[bookbrainz-data-js] 14kellnerd opened pull request #314 (03master…repeatable-imports): Repeatable imports https://github.com/metabrainz/bookbrainz-data-j...

My last PR for GSoC, repeated imports can now "overwrite" pending entities :)

kellnerd: congrats on the work! 🥳