#musicbrainz-devel

but your ticket said replicated

anyway, it uncovered a big problem

which was that running the upgrade on slaves would create some foreign keys and stuff too

so i had to rip it all apart

ocharles: printf "Foo\nBar\n" | sed 's/^/echo /' | sh

its said "replicated slave"

warp: oh, heh

not replication triggers.

ocharles: yep, that was the problem I wanted you to fix.

ah, right

was there something else changed?

i guess i confused myself, but ended up fixing the right problem :)

yes. :)

ok, I'm pushing seach indexes right now.

we did a full master import with this on hobbes earlier

next I will do a slave test to see if it works.

adn upgrade.sh ran as RT_MASTER?

no, only tested as RT_STANDALONE

ok, I will do a slave and if I have time a master.

first some fudz

but RT_MASTER just adds replication triggers to all that, so I would hope that's in the same area of testing

RT_MASTER does other things too. those code paths should be retested too

do you have an estimate on how long does admin/sql/updates/20120314-add-tracknumber.sql take to run?

less than 5 minutes.

cool, thanks

depends on the speed of your disks.

was about 3 minutes on my SSD mac

ocharles, nikki: alright. all my stuff is done again. pushed to schema-change-2012-05-15 .

luks: it's just over 5 on hobbes

and hobbes only has 7.2k drives

5.4k only!

it's with great pleasure I announce the latest in our line of Cover Art Archive projects

Ladies and gentlemen, introducing...

lol. great naming.

have you seen the latest bug on CAA?

i don't know, which bug are you thinking of?

CAA-22 IRC

yes, that's why I wrote this

at least for the cleanup part

ok

do we know why it happens?

still need to fix the trigger to not create events that don't need any work

yep

k

but before I run this, it'd be nice to know if the IA have back ups :)

I'm running find-empty piped into s3cmd ls for now

”

that'd be the one :)

ocharles, warp: is the schema-change branch updated with all the changes we believe we need for DB changes?

I want to run upgrade.sh in my sandboxes

question poised to the archive.

Let me double check

warp: all your stuff pushed?

warp: prod

schema-changes-2012-05-15 pushed, a850ea3b8cac610

I'm ready to go with that branch.

warp???

pinging him won't help :)

can you send him a txt please?

+31 is NL right?

yeah

sent

thx

yes

:)

you can also ring me on skype

ok

is all your stuff pushed?

yes.

only a few small fixes comitted today, but they're all merged into the schema-change-2012-05-15 branch and pushed.

ok, master Vm created.

now duplicating a slave copy

ruaok: can you ask the IA what they want from a cleanup?

http://archive.org/details/mbid-e70f3bd2-cf74-4... seems to be as much as I can do

hmm, that doesn't help you actually

basically, I can remove all files from a bucket but can't remove the bucket. is that enough?

if the buckets should be removed, just make a list of buckets to nuke.

we can then pass that over and have sam do it.

slave test fired off

ocharles: http://wiki.postgresql.org/wiki/Binary_Replicat...

which steps do I need to apply to totoro's config?

"Setting Up Archiving On The Master"

We use nfs here, but the rsync method looks like it should be good

well, but we're not actually going to deploy it just yet.

right, but I don't know if you can turn it on without a restart

one moment

fine with a restart.

since we have down time anyway.

its just that the slave isn't there yet

i mean when we decide to turn this on

I think we might need to start shipping logs to the slave, and then just delete them on the slave

ah

the slave isn't setup at all.

just a blank machine with a blank install.

ok

turns out you can change archive_command without a restart

can you have a null archive_command?

'Setting archive_command to a command that does nothing but return true, e.g. /bin/true (REM on Windows), effectively disables archiving, but also breaks the chain of WAL files needed for archive recovery, so it should only be used in unusual circumstances.'

so, no?

so yes :)

my feeling is that we should just skip it for now

wal_level = hot_standby; archive_mode = on; archive_command = /bin/true

breaking the write-ahead log seems like a bad idea

breaks the chain

that's not a problem if there's no chain to break

tbh, I think I would be more comfortable with not doing this now.

doing a single DB restart means 30 seconds of downtime. I can deal with that in the future.

it's literally what I pasted above. a restart is not just 30 seconds of downtime, it will clear all your indexes too

but it's your call

clear indexes?

i can appreciate limiting the amount of work we do

from ram

yeah, thats not a big deal.

lets focus on the giant pile of stuff we have to get right.

and not make it bigger.

slave test passed.

search indexes are in place.

I'm going to run to the shop to pick up something for dinner

won't be more than a few minutes

k

have all the tests passed?

yea

shall I merge to master and get jenkins to see what it thinks?

plz

329 files changed, 14685 insertions(+), 1788 deletions(-)

and I'll do another slave test after that.

yay, merging to master.

is hobbes busy?

ci.musicbrainz.org isn't loading for me

I'm not using hobbes

ok, build kicked off

Project musicbrainz-server_master build #189: FAILURE in 9 min 47 sec: http://ci.musicbrainz.org/job/musicbrainz-serve...

crap

meh

right, times to be fixings

odd, the merge seems to have reverted a change somewhere...

which is a bit unsettling

ok, i think that's me just reading things wrong

release time is here!

:-(

isn't that due to the other failures?+

quite possibly.