#metabrainz

/

      • bitmap
        yvanzo: it's on paco now
      • 2020-05-19 14043, 2020

      • yvanzo
        (I reproduced the shared memory issue using test report)
      • 2020-05-19 14046, 2020

      • yvanzo
        thanks
      • 2020-05-19 14035, 2020

      • ruaok
        legoktm: do you have any ideas what we should do about our IP block being blocked by wikimedia?
      • 2020-05-19 14008, 2020

      • legoktm
        Yes
      • 2020-05-19 14026, 2020

      • legoktm
        Is it a dedicated IP? We can apply for a block exemption
      • 2020-05-19 14040, 2020

      • ruaok
        how do we do that?
      • 2020-05-19 14044, 2020

      • ruaok
        yes , it is.
      • 2020-05-19 14059, 2020

      • ruaok
        bitmap: what was the external IP of the server that runs the bot?
      • 2020-05-19 14030, 2020

      • bitmap
        the error they returned says it's 138.201.203.23
      • 2020-05-19 14029, 2020

      • ruaok
        yep, makes sense. zappa.
      • 2020-05-19 14052, 2020

      • legoktm
        The bot is logged it with an account, right? What's the username?
      • 2020-05-19 14032, 2020

      • bitmap
        MineoBot I think?
      • 2020-05-19 14044, 2020

      • ruaok
        that I do not know.
      • 2020-05-19 14055, 2020

      • ruaok
        sounds right.
      • 2020-05-19 14008, 2020

      • legoktm
        https://www.wikidata.org/wiki/Special:Contributio… looks like it last edited in October
      • 2020-05-19 14048, 2020

      • legoktm
        I'm going to send an email to the Wikimedia stewards, who should I cc on it?
      • 2020-05-19 14000, 2020

      • ruaok
        reosarevok@meb , please
      • 2020-05-19 14026, 2020

      • yvanzo
        bitmap: pgbouncer is actually running from the same container as postgres, doesn't seem possible to use PAUSE/RESUME at all.
      • 2020-05-19 14058, 2020

      • bitmap
        how come?
      • 2020-05-19 14008, 2020

      • yvanzo
        or am I missing something?
      • 2020-05-19 14018, 2020

      • bitmap
        you have to connect to the pgbouncer database on whatever port pgbouncer is listening on (6899 internally, not sure what the external one is)
      • 2020-05-19 14019, 2020

      • bitmap
        pg & pgbouncer run inside the same container but listen on different ports
      • 2020-05-19 14054, 2020

      • bitmap
        after the PAUSE you'd probably want to pg_ctl restart and then RESUME asap
      • 2020-05-19 14006, 2020

      • legoktm
        ruaok, bitmap: here's what I sent for reference: https://paste.centos.org/view/raw/61f66af0
      • 2020-05-19 14035, 2020

      • yvanzo
        bitmap: I need to restart the container, not just pg_ctl.
      • 2020-05-19 14037, 2020

      • ruaok
        <3 thank you!
      • 2020-05-19 14004, 2020

      • reosarevok
        Thanks legoktm!
      • 2020-05-19 14054, 2020

      • yvanzo
        Thanks!
      • 2020-05-19 14026, 2020

      • bitmap
        oh, sorry
      • 2020-05-19 14034, 2020

      • bitmap
        I keep thinking we're just restarting it for the postgresql.conf changes
      • 2020-05-19 14015, 2020

      • legoktm
        np :)
      • 2020-05-19 14029, 2020

      • yvanzo
        bitmap: no problem, I wasn't sure to understand it all :)
      • 2020-05-19 14013, 2020

      • bitmap
        (that is an argument for running pgbouncer in a separate container, though)
      • 2020-05-19 14018, 2020

      • Chinmay3199 joined the channel
      • 2020-05-19 14026, 2020

      • yvanzo
        should I stop sentry first?
      • 2020-05-19 14015, 2020

      • yvanzo
        (I will wait for json-dump that should run in 2min)
      • 2020-05-19 14007, 2020

      • bitmap
        should be fine to leave sentry running, might log some internal errors but
      • 2020-05-19 14017, 2020

      • bitmap
        the other option for minimizing downtime is doing a failover to pink, but that hasn't been tested in production either, heh
      • 2020-05-19 14008, 2020

      • bitmap
        that's all handled by repmgr and there are tests in the docker-postgres-cluster repo to show it works, but still a bit scary
      • 2020-05-19 14039, 2020

      • MajorLurker joined the channel
      • 2020-05-19 14057, 2020

      • yvanzo
        we should definitely try to put failover in place but that might not be the right time to test it :D
      • 2020-05-19 14009, 2020

      • ruaok
      • 2020-05-19 14010, 2020

      • BrainzBot
        LB-599: lastest_listened_ts is not longer used
      • 2020-05-19 14003, 2020

      • iliekcomputers
        Phew
      • 2020-05-19 14019, 2020

      • iliekcomputers
        I was worried my client would break
      • 2020-05-19 14023, 2020

      • iliekcomputers
        Thanks for digging into it, ruaok
      • 2020-05-19 14053, 2020

      • ruaok
        well, it caused the integration tests to fail. go us. :)
      • 2020-05-19 14003, 2020

      • ruaok
        said ironically and not.
      • 2020-05-19 14032, 2020

      • ruaok
        I <3 real time continuous aggregates.
      • 2020-05-19 14017, 2020

      • ruaok
      • 2020-05-19 14032, 2020

      • ruaok
        now we have fast access to the min and max timestamps.
      • 2020-05-19 14017, 2020

      • d4rkie joined the channel
      • 2020-05-19 14059, 2020

      • yvanzo
        bitmap: that worked, running DuplicateArtist report no longer triggers this error.
      • 2020-05-19 14012, 2020

      • Nyanko-sensei has quit
      • 2020-05-19 14009, 2020

      • yvanzo
        however, there is an ISE related to musicbrainz_collate at https://test.musicbrainz.org/recording/fd715073-2… (doesn't happen in production)
      • 2020-05-19 14002, 2020

      • bitmap
        oh yes, I ran the pg12 upgrade scripts on the musicbrainz_schema_change db, but the test containers haven't been updated yet
      • 2020-05-19 14024, 2020

      • bitmap
        was rebuilding new images yesterday but I think a ' in the commit message broke something, lol https://ci.metabrainz.org/job/MusicBrainz%20Serve…
      • 2020-05-19 14015, 2020

      • yvanzo
        should we proceed with other postgres instances now?
      • 2020-05-19 14014, 2020

      • yvanzo
        or maybe we should schedule it for tomorrow?
      • 2020-05-19 14013, 2020

      • bitmap
        we could do it on pink today & floyd tomorrow
      • 2020-05-19 14056, 2020

      • yvanzo
        ok, that would unlock reports at least
      • 2020-05-19 14032, 2020

      • bitmap
        reports are run on floyd though since they write to the db
      • 2020-05-19 14010, 2020

      • yvanzo
        :(
      • 2020-05-19 14022, 2020

      • bitmap
        we should fix mbs to switch to pink when in DB_READ_ONLY mode, and not break when that's the case
      • 2020-05-19 14009, 2020

      • bitmap
        which might just be preventing people from logging in since it errors on trying to update last_login_date :)
      • 2020-05-19 14053, 2020

      • legoktm
        bitmap: the block exemption is in place now, so the bot should be able to edit now :)
      • 2020-05-19 14010, 2020

      • bitmap
        wow that was fast :) thanks!
      • 2020-05-19 14026, 2020

      • ruaok
        very cool, thanks!
      • 2020-05-19 14048, 2020

      • bitmap
        yes I don't see any errors in the wikidata-bot logs anymore 👍
      • 2020-05-19 14056, 2020

      • ruaok
        woo!
      • 2020-05-19 14042, 2020

      • legoktm
        :D
      • 2020-05-19 14045, 2020

      • legoktm
      • 2020-05-19 14026, 2020

      • bitmap looks at fixing its editing on the MB side, looks like a consul key is missing
      • 2020-05-19 14052, 2020

      • bitmap
        zas: I pushed an update to a consul config in docker-server-configs, but it doesn't seem to be getting into consul, not sure why
      • 2020-05-19 14044, 2020

      • zas
        bitmap: which update?
      • 2020-05-19 14057, 2020

      • bitmap
        when I set "down": "yes" in the upstream_services config yesterday it didn't work either, maybe the same issue
      • 2020-05-19 14007, 2020

      • bitmap
      • 2020-05-19 14021, 2020

      • bitmap
        git2consul down/broken maybe?
      • 2020-05-19 14011, 2020

      • zas
        restarting it
      • 2020-05-19 14005, 2020

      • bitmap
        yeah, the key magically appeared now
      • 2020-05-19 14012, 2020

      • ZaphodBeeblebrox joined the channel
      • 2020-05-19 14012, 2020

      • ZaphodBeeblebrox has quit
      • 2020-05-19 14012, 2020

      • ZaphodBeeblebrox joined the channel
      • 2020-05-19 14034, 2020

      • CatQuest has quit
      • 2020-05-19 14040, 2020

      • v6lur_ has quit
      • 2020-05-19 14048, 2020

      • rdswift has quit
      • 2020-05-19 14057, 2020

      • Chinmay3199 has quit
      • 2020-05-19 14003, 2020

      • rdswift joined the channel
      • 2020-05-19 14013, 2020

      • BrainzGit
        [musicbrainz-server] mwiencek opened pull request #1521 (production…db-read-only-db): Default to READONLY database for DB_READ_ONLY https://github.com/metabrainz/musicbrainz-server/…
      • 2020-05-19 14036, 2020

      • bitmap
        yvanzo: ^ we can deploy that before restarting floyd, then MB can stay up in read-only mode with all connections going to pink
      • 2020-05-19 14013, 2020

      • bitmap
        also turns out there isn't an issue with last_login_date, it already skips updating that with DB_READ_ONLY set
      • 2020-05-19 14048, 2020

      • d4rkie has quit
      • 2020-05-19 14037, 2020

      • Nyanko-sensei joined the channel