#musicbrainz-devel

I'm going to go down the no email means it's all ok route

k

oh, ocharles can I borrow you for 3 min on a different topic? ping me when ready....

djce: fire away

ok. Just wanted to briefly discuss http://tickets.musicbrainz.org/browse/MBH-153

I updated the spec on that ticket

and I think it'll mainly (probably only) affect the MBS codebase in fact.

yay, md5sums

So some combination of a FYI, and a "please shout if you see a problem",

and my next task is to estimate how long that'll take

Yea, it looks like it's just MBS

so I can keep my boss updated as to how long I'm bunking off BBC work to do MBH instead :-)

:)

I'm still unclear as to where I'll be testing this, alas.

Might have to resurrect replication packet emitting on hobbes.

but I have no access to any MB slaves :-(

hobbes is pretty much ready for that I'd think

(well.... hmm... I could use pretty much any machine in fact if I just mock the bit where it loads the packet into the database I suppose).

ok. Thanks. that's all for now, just wanted to keep you informed mainly.

(damn, 10 sec over my 3 minute chat estimate ;-)

lol

ISE on the release editor, MBS-3165 if someone wants to capture stack traces etc from the error log.

Sigh, we're getting 502s again it seems

goodmorning!

cjk32: regarding the session timeout, it's probably better to submit what you have, and add an editnote saying you will add the remaining discs later.

still, it shoudln't "internal server error" ofcourse

PostgreSQL indexing and plperl just blew me away

I can find all edit relationship edits that use a certain relationship type in....

0.36ms

planning to use plperl in MB?

only for this json searching

i can't see any other way to query it fast otherwise

(http://tickets.musicbrainz.org/browse/MBS-2676)

Does the android app use /ws/2 or /ws/1?

If I search for release 'Aerial' and select the Kate Bush one, it only lists 5 matches, all for 'A Sky of Honey'. None for the other disc in the release

The matches all have 16 tracks, whereas in reality it's two discs, one with 7, the other with 9.

warp: http://codereview.musicbrainz.org/r/1450/diff/#... could you explain the last hunk?

also, luks any idea on http://tickets.musicbrainz.org/browse/MBS-3162?

echo "SELECT plainto_tsquery('mb_simple', '☺')" | ./admin/psql READWRITE just gives "NOTICE: text-search query contains only stop words or doesn't contain lexemes, ignored"

ocharles: when in the release editor you add a track to a VA release, the artist credit for that track will be empty.

ocharles: if you submit that you will get a warning that the track needs an artist

ocharles: but without these changes you can never submit a new artist, because while iterating over each credit 'current' will be falsy (because it is empty) for each credit.

hrm

ocharles: not very easy to fix, we always had that problem (mainly with !!! :))

well, I don't really get it but if it works the rest looks ok :)

luks: yea

I wondered if there was a way to just do a direct text comparison as well

ocharles: hrm, perhaps it would be cleaner and clearer to just compare to.names.length with from.names.length

well, you can do "... OR artist_name.name = ?" :)

ocharles: before the $.each() over to.names.

luks: yea, that's probably what I will be doing

but it definitely doesn't look like it needs to get fixed in 2011-08-08

but that should presumably rank at the very top too

well, it's in there :)

warp put it there

I asked for other work yesterday, and didn't really get any

yea

it's a very minor issue

MB had it since the beginning

Everything in my milestone that I can actually work is pretty minor, other than http://tickets.musicbrainz.org/browse/MBS-3017 which I still need to find more info on and http://tickets.musicbrainz.org/browse/MBS-2676 which needs me to finish off with fresh eyes tomorrow

sure, that doesn't make it any less serious :)

if we lived with it for 10 years, it can wait if there is more important work (which I'm sure there is)

ocharles: all ModBot errors fixed? :)

all known ones

not that it's too easy to filter a list of 700 tickets I have assigned to me :P

luks: ofcourse to some extent minor/critical is subjective. I have music my this particular artist in /mnt/music/tagged, so I'd more likely to be hit by it than most people :)

ocharles: go by the number of votes

http://tickets.musicbrainz.org/secure/IssueNavi... is what I otherwise work by

but I get told off for not working off my CI list :)

working on that search thing really seems like a waste of time to me

yep

or just adding to the list of exceptions, which currently contains only !!!

add

we only do that for indexed search atm, not direct searches

oh, I thought the direct search also used it

but it really should be much work to allow the direct search to do exact matches as well.

+not

ocharles: if you don't feel like working on it, assign it to me and I'll look into it.

i don't mind working on it, I just kinda agree that we have more pressing things

so I'm working on http://tickets.musicbrainz.org/browse/MBS-3127 instead

you cannot assign this artist to tracks and recordings, etc.. at all now. that seems pretty important.

right, time to wander home

ruaok: we're getting a lot of 502 errors the last few days

ah, ok

not sure what to do about it.

heard it on twitter, actually. ;)

yea, I saw that

but also saw it mentioned in IRC the day before

I thought it was due to having an open psql shell to totoro, but that's apparently not it

why would that do it?

connection count

the backend servers are all happy. nothing out of sorts when what I saw.

our connection count is nuts.

before I've accidently left a transaction open too, and that wrecked a bit of havoc

but that wasn't the problem this time (there was no transaction open)

*that* will do it.

but once you start hitting your connection limit, closing one connection won't help.

right

closing a 100 connections might have a positive effect.

it was only a wild guess, I doubt that's what it was

I think I will restart the front end proxy and see what that will do.

I think it was just coincidence 502s stopped briefly when I quit the shell

to diagnose 502s, here is your process:

1. Are the back ends happy? if not, the front end requests are piling up for some reason. thus 502s.

how do I measure happiness?

tail the logs?

you look at server loads at the first clue.

ok

then you start looking at error/access logs

but the mrtg graphs are your first line of insight.

if the back ends are happy its likely that the proxy itself is having issues.

which is kinda hard to diagose at times.

but looking at the logs is the first thing to do.

other than just kick it and see what happens :)

ruaok: server load average should be < # of cores in the system?

ocharles: that used to be my preferred way of doing things, but rather ineffective. ;)

heh

warp: yeah.

given that, astro and pingu sitting at load 6 is cake.

totoro at 2... he's nearly napping he's so bored.

oh, talking of totoro, poke me when you're free later - need to discuss something with you

totoro napping seems so fitting :P

haha

ocharles: after coffee. ;)

2011/08/04 17:15:24 [warn] 26957#0: *1590920301 an upstream response is buffered to a temporary file

sure

we're seeing too many of those.

did we just push out something that made the size of the /ws/1/release responses bigger?

ok, I'll keep tailing the logs for 502s.