zas: I'd like to move the postgres instance on williams to paco, rn williams is a bit too overloaded with json dumps, search index dumps, barman backups, jenkins builds, and slow disks
supersandro2000 has quit
supersandro20006 joined the channel
can barely run a 200GB rsync locally without waiting 3 hours :/
lmk if there's a better spot but looks like paco has enough space and is relatively underused
hey. so, pg12 coming tomorrow. I failed to put that on our radar.
which means that PG 9.5 goes down and then comes back up and migrated as far as CB, LB and MeB are concerned.
iliekcomputers
Hmm, I saw the PR.
ruaok
I *think* it will be a non-event.
iliekcomputers
Does the password thing mean that we'll have to release new versions of LB and CB
?
Or is that for dev only?
ruaok
unsure. let me see about passwords in production.
but I think that thing for us to do is to put passwords on all PG accounts going forward. that should avoid trouble.
the password is specified in consul_config.py.ctmpl, so it should be fine.
things even seem to work with the pg9.5 postgresql client inside docker images.
however, the messybrainz tests are all failing and they are setup differently than the LB tests.
can you please take a look and see how to set the password for the messybrainz tests?
iliekcomputers
Yeah, I can look at that.
ruaok
thx.
once we get LB tucked away we need to look at MeB and CB.
iliekcomputers
I think it might be better if you take MeB, I'll take LB and then we'll circle back to CB
ruaok
sure, that works.
iliekcomputers
LB is mostly done anyways, thanks to your PR.
ruaok
on Meb as we speak.
and CB is likely to be very similar to MeB.
CatQuest
Gratulerer med Dagen!
🇳🇴
Gazooo has quit
Gazooo joined the channel
ruaok
all meb tests pass with pg12.3
iliekcomputers: should I look at how to install postgres-client-12 or have you looked into that already?
iliekcomputers
i'm looking at the tests
go ahead
ruaok
ok
iliekcomputers
that way i can cargo-cult :D
ruaok
obvs
iliekcomputers
8 messybrainz tests failing...
they seem to be actual issues
i don't think we run any of this code at all in prod
it's from kartikeya's gsoc project
ruaok
oh, that is why absolutely nothing made sense yesterday.
messybrainz may have one problem we should look at.
pg12 requires function names to be fully qualified. and when I import the msb data into timescale, which is pg12, there are problems resolving the array_agg function.
iliekcomputers
do we use array_agg somewhere in prod
ruaok
yes. array_agg is used by artist_sort. And neither calls to array_sort are qualified.
db/artist.py uses array_sort.
doesn't look like a huge problem
iliekcomputers
oh yeah, artist.py is the 8 failing tests
ruaok
but that may all be unused code from katikeyas project, I would guess.
quite frankly we could back that code out if you prefer.
the mapping is the way forward for the time being.
if the other approach gets picked up again, we can cobble together code from it.
iliekcomputers
yeah, make sense.
i'll try to fix the array_agg and see what happens