#musicbrainz-devel

/

      • kepstin-laptop notes that the X240 has a squished keyboard with a firmware issue that turns the end key into insert when you enable function-lock (i.e. so your F* keys are F* keys instead of brightness, etc.)
      • kepstin-laptop
        since insert is Fn+End, and the Fn-lock swaps all of the top-row keys with special functions :/
      • whoa, that's weird. on the new X1 carbon they moved the `~ key to between the right alt and ctrl at the bottom
      • and escape went where it used to be :/
      • (this video appears to be using the US layout)
      • adhawkins
        ocharles: Figured you'd have seen it already. Made me chuckle :D
      • Server still having problems? Any ETA? (Not hassling, just wondering when I should next check it)
      • chirlu`
        Reminds me, there is a small Haskell program that I need to finish.
      • CallerNo6 , when he finally gets the thinkpad he's lusting after, will probably glue a cover over the touchpad/trackpad/thingy. hate those things.
      • kepstin-laptop
        CallerNo6: easier to disable it in software, looks better too.
      • CallerNo6
        But what if I glue somethng /useful/ over it? Picture frame? Big Gulp holder?
      • kepstin-laptop
        and if you get a late-2013/2014 model, you need the touchpad available because its touch-sensitivity is used to tell which trackpoint button you're pressing.
      • (this means anything with the last two digits of the model number >30)
      • CallerNo6
        fortunately, a new one isn't in the budget anyway
      • voiceinsideyou joined the channel
      • voiceinsideyou2 joined the channel
      • rvedotrc joined the channel
      • misterswag joined the channel
      • djinni` joined the channel
      • zas__
        bitmap, luks_, Mineo & others : what do you think about http://tickets.musicbrainz.org/browse/PICARD-408 ?
      • ijabz joined the channel
      • bitmap
        zas__: can you check if the left pane is clipped (has scrollable contents) and just resize the window/pane accordingly?
      • I'm not too fond of creating our own checkbox widget
      • zas__
        bitmap: the left pane (tree) isn't scrollable atm
      • bitmap: i'm not fond of new checkbox widget too, and perhaps some other widget has the same behavior...
      • bitmap
        hm, it's scrollable on osx
      • in any case there might be a way to compare the width of the pane to the width of the inner tree, but I'm not too sure
      • chirlu`
        reoafk: Mineo: Squashed the bugs related to the statistics pages, feel free to test at http://chirlu.mbsandbox.org/
      • ruaok
        bang!
      • meeting time.
      • ianmcorvidae
        hello
      • ruaok
        is there a postgres doctor in the house? ;-)
      • morning ianmcorvidae
      • fractalizator2 joined the channel
      • ocharles and I made nearly no progress on totoro.
      • I've got a script that kills stucks procs, which is keeping things sane.
      • we've tried to up shared_buffers. no effect.
      • ianmcorvidae
        yeah. I hear there was some vacuuming that found weirdnesses in the edit table too, or?
      • ruaok
        I just upped effective_cache_size to 20GB and that relieved some pressure, but that artist edits still don't load.
      • yes, we got two error messages, but nothing else
      • I'm flat out of ideas, sadly.
      • ianmcorvidae
        I wonder if we've done the same for edit_artist
      • but yeah, I'm not sure either
      • ruaok
        have not.
      • probably a good idea.
      • LordSputnik joined the channel
      • shall I start that?
      • ianmcorvidae
        (it's mostly been artists that have been the issue, so)
      • well, if it's fine for now we may as well do our meeting at least :P
      • ruaok
        ok.
      • besides travel and new years and pg issues, I've got nothing really new.
      • anyone have anything to discuss?
      • Freso likes that ruaok and ianmcorvidae's nicks both show up as cyan - also, is here
      • ruaok kicks off the vacuum
      • Freso
        nikki: Ping?
      • ianmcorvidae
        not much here either; I took wednesday off and worked on some various things around it
      • did at least get one long-lived ticket up for review
      • Freso
        I guess nikki's the only one with something on the agenda. She was around a few minutes ago.
      • ruaok
        ianmcorvidae: did you test increasing the rate limit and seeing if it had an effect on rika?
      • ianmcorvidae
        I've increased it, but haven't seen what it's done for rika
      • nikki, reoafk, bitmap, chirlu might be able to weigh in if it's been better the last couple days or not
      • ruaok
        another thought would be to pay attention to the peaks and troughs in our 95th graph.
      • rika should be nicer to use in a trough, than a peak
      • bitmap
        rika was fast for me when I was using it yesterday
      • ruaok
        if there is anything to the bandwidth limiting idea.
      • djinni` joined the channel
      • derwin
        stop typoing "riak"
      • bitmap
        was even able to submit a 40 disc release on it :)
      • chirlu`
        There is a delay when typing on rika, but I would consider it normal.
      • Can't compare to previous weeks, though.
      • reodroid joined the channel
      • ruaok
        hmm.
      • ianmcorvidae
        okay. well, we'll see if the small improvements stick
      • Freso
        Seems snappier to me too right now.
      • bitmap
        the ssh delay is better than it usually is right now
      • ruaok
        well, if anyone encounters trouble, note where in the 95th graph we are
      • ianmcorvidae
        it's always been slightly slow, intentionally, but usably so is a change
      • ruaok
        http://stats.musicbrainz.org/mrtg/95th-percenti...
      • interestingly enough we're headed to a peak now and its doing well.
      • what else can we discuss on rika?
      • watch and observe, what else?
      • ianmcorvidae
        I'm not sure there's much else that'd be part of this topic
      • I need to dist-upgrade it at some point but that's probably unrelated
      • ruaok
        k.
      • any other topics?
      • Freso
        What's for dinner?
      • Freso hides
      • ruaok
        burger wraps here. :)
      • ianmcorvidae
        a few hours away, here :P
      • but nothing more from me, I don't think
      • ruaok
        well, if no one has anything else to discuss, then I think I'd love to chat with ianmcorvidae and ocharles about totoro.
      • see if the three of us (and anyone else who'd care to participate) can think up new things to check out.
      • ianmcorvidae
        how's that edit_artist vacuum doing?
      • ruaok is looking forward to a more action packed meeting next week.
      • and the main thing that comes to mind is why the MBS timeout isn't working
      • ruaok
        still going. load 13, 6-10MB/s io
      • let's close the general meeting, then.
      • ianmcorvidae
        we fix that and it still 504s, but it doesn't start slowly breaking totoro in the process
      • ruaok
        sorry for not delving into DRs this week, but the sick DB server is important for us to pay attention to.
      • thanks everyone!
      • /bang!
      • reodroid
        heh that was fast :)
      • nikki
        oh
      • ruaok
        nikki: ?
      • ianmcorvidae: yes, the timeout killer is odd -- ocharles tried it a couple of different ways and it worked.
      • except in production
      • ruaok shrugs
      • nikki
        I managed to miss the entire meeting :P
      • ianmcorvidae
        one question I had was what our old version of pgbouncer was
      • since with the dist-upgrade we switched to a package manager-installed one from a hand-compiled one
      • ocharles: you'd seem to be the one who might know on that count
      • Freso
        nikki: I did try to ping you. :)
      • ruaok
        ianmcorvidae: ah, that is an interesting idea.
      • which fits with your observation that something changed at dist-upgrade time
      • also, jozo asked if the collate problem that prevented the replication packets from applying could hamper our server.
      • ianmcorvidae
        yeah -- though I will stress again that it wasn't anything near this bad until last night, of course
      • ruaok
        do we need to rebuild those indexes?
      • ianmcorvidae: yes, but I've seen this pattern before.
      • things a slightly elevated and then you hit some inflection point and BAM, things go south.
      • ianmcorvidae
        we might need to because postgresql-musicbrainz-collate's behavior has changed, but we should actually update it first
      • ruaok
        I've kept forgetting to mention this for the past few meetings. this had been making me uneasy, TBH
      • ianmcorvidae
        I think ollie got the package updated though, so if you've done some provisions we may be fine
      • (on the collate issue)
      • ruaok
        we did a provision today to up the shared_buffers.
      • ianmcorvidae
        any chance you remember if it updated a postgresql-musicbrainz-collate package in the mix there?
      • ruaok
        that reminds me, I need to back-propagate the effective_cache_size change
      • not i.
      • ocharles was driving.
      • should we pre-emptively hand-compile a pg-bouncer?
      • ianmcorvidae
        without knowing what version the old one was I don't know that it'd help
      • I don't know if we upgraded or downgraded or neither :)
      • ruaok
        where would that have gotten installed?
      • into the postgres install dir?
      • or elsewhere?
      • maybe its still around.
      • ianmcorvidae
        it installed to /usr/local/pgbouncer or so, but you can't run the binaries
      • because of missing libraries from the pre-dist-upgrade server
      • ruaok
        strings?
      • ianmcorvidae
        hm, maybe
      • ruaok
        vacuum of edit_artist is still going
      • ianmcorvidae
        1.5.2, current version installed is 1.5.4-3.pgdg12.4+1
      • ruaok
        yay.
      • ianmcorvidae wonders if I can find a pgbouncer changelog
      • I'd say lets compile 1.5.2 and try to use it and see what happens.
      • along these lines, of course: http://jo-kes.blogspot.com.es/2007/11/three-eng...
      • ianmcorvidae also looks into pgbouncer configuration options that might help us
      • ianmcorvidae
        hm
      • only the 'dangerous timeouts' section looks like it could help, so that's less than ideal
      • ruaok
        huh.
      • the vacuum is still going and yet things are really calm now.