#metabrainz

/

      • D4RK-PH0_ has quit
      • 2021-01-20 02031, 2021

      • Nyanko-sensei joined the channel
      • 2021-01-20 02018, 2021

      • Etua1 joined the channel
      • 2021-01-20 02042, 2021

      • Etua has quit
      • 2021-01-20 02043, 2021

      • Etua1 is now known as Etua
      • 2021-01-20 02020, 2021

      • Etua has quit
      • 2021-01-20 02049, 2021

      • sumedh joined the channel
      • 2021-01-20 02044, 2021

      • ephemer0l_ has quit
      • 2021-01-20 02022, 2021

      • sumedh has quit
      • 2021-01-20 02038, 2021

      • yvanzo
        mo´´in´
      • 2021-01-20 02006, 2021

      • zas
        moooin
      • 2021-01-20 02025, 2021

      • reosarevok
        niom!
      • 2021-01-20 02000, 2021

      • ephemer0l joined the channel
      • 2021-01-20 02003, 2021

      • zas
        bitmap: I just truncated postgres-floyd log file (it was 252Gb big), daemon config was changed but daemon never restarted. There's a way to do that at runtime somehow, but I'd prefer you are around if we try that....
      • 2021-01-20 02015, 2021

      • diru1100 joined the channel
      • 2021-01-20 02027, 2021

      • diru1100
        Good afternoon!!
      • 2021-01-20 02018, 2021

      • diru1100
        Just completed FOSS exam, literally wrote everything related to MetaBrainz.
      • 2021-01-20 02024, 2021

      • shivam-kapila
        interesting subject
      • 2021-01-20 02026, 2021

      • shivam-kapila
        jenkins appears to be down. I hope its known
      • 2021-01-20 02044, 2021

      • zas
        bitmap, ruaok, alastairp: we still have disk space issues on williams
      • 2021-01-20 02042, 2021

      • zas
        shivam-kapila: it should be ok now, it was me, I had to reload docker daemon config on williams and jenkins issue is likely related
      • 2021-01-20 02050, 2021

      • zas
        can you confirm it works?
      • 2021-01-20 02031, 2021

      • shivam-kapila
        zas: yeah its working now. thanks
      • 2021-01-20 02002, 2021

      • alastairp
        zas: I freed up less than 10gb
      • 2021-01-20 02058, 2021

      • alastairp
        on Friday I will work on the automatic cleanup tasks that I mentioned last time, and I will move to cage, we should check if there are extra wal files again like bitmap found last week. I don't know where they are
      • 2021-01-20 02038, 2021

      • d4rkie has quit
      • 2021-01-20 02010, 2021

      • Etua joined the channel
      • 2021-01-20 02047, 2021

      • zas
        there are in /var/lib/docker/volumes/barman-backups/_data/floyd 427Gb
      • 2021-01-20 02054, 2021

      • zas
        bitmap: ^^
      • 2021-01-20 02016, 2021

      • loujine has quit
      • 2021-01-20 02057, 2021

      • Etua1 joined the channel
      • 2021-01-20 02014, 2021

      • loujine joined the channel
      • 2021-01-20 02037, 2021

      • Etua has quit
      • 2021-01-20 02037, 2021

      • Etua1 is now known as Etua
      • 2021-01-20 02052, 2021

      • d4rkie joined the channel
      • 2021-01-20 02004, 2021

      • ruaok
        moooin!
      • 2021-01-20 02021, 2021

      • ruaok
        zas: that file is on williams? that would explain the diskspace problems. :)
      • 2021-01-20 02000, 2021

      • zas
        ruaok: that's multiple files, that's barman collecting wals from postgres on floyd to back up main db, but for some reason they sometimes accumulate. Issue was identified a while ago, but we don't know yet what causes it. bitmap could prolly tell more about this.
      • 2021-01-20 02033, 2021

      • zas
        but that's not the sole disk space user on williams, jenkins is not cleaning all its mess
      • 2021-01-20 02045, 2021

      • zas
        (alastairp will take care of this soon)
      • 2021-01-20 02059, 2021

      • zas
        and ftp uses almost 700Gb
      • 2021-01-20 02027, 2021

      • alastairp
        but jenkins runs over by ~10gb every few weeks, with is small fry compared to stuck barman backups and ftp
      • 2021-01-20 02037, 2021

      • zas
        mainly because of listenbrainz dumps: 439gb atm
      • 2021-01-20 02018, 2021

      • zas
        alastairp: yup
      • 2021-01-20 02001, 2021

      • alastairp
        did I recall ruaok mentioning that we could use the storage box for ftp?
      • 2021-01-20 02017, 2021

      • zas
        yes, we could for ftp at least
      • 2021-01-20 02012, 2021

      • zas
        ah, listenbrainz is 539gb (not 439) atm
      • 2021-01-20 02020, 2021

      • zas
        mainly in /home/ftp/ftp/root/pub/musicbrainz/listenbrainz/fullexport
      • 2021-01-20 02022, 2021

      • ruaok
        an the FTP situation is only going to get worse as we publish more data sets.
      • 2021-01-20 02052, 2021

      • ruaok
        for hetzner cloud there are attachable storage devices that perform and work much better than storage boxes.
      • 2021-01-20 02018, 2021

      • ruaok
        I wonder if we can attach these to bare metal servers yet. that would be the ideal method for moving forward on this. let me look.
      • 2021-01-20 02025, 2021

      • antlarr has quit
      • 2021-01-20 02039, 2021

      • antlarr joined the channel
      • 2021-01-20 02007, 2021

      • ruaok
      • 2021-01-20 02032, 2021

      • ruaok
        2TB for 11.78 €? that works.
      • 2021-01-20 02027, 2021

      • ruaok
        doesn't seem like those can be mounted into a FS and are used on a "sync" basis.
      • 2021-01-20 02048, 2021

      • reosarevok
        bitmap, yvanzo: what do you think of https://tickets.metabrainz.org/browse/MBS-11339 ? It seems sensible to me, but if we do should we still keep a redirect for those /artist/gid/credit/id links in case someone is using them externally for some reason? I doubt it, but :)
      • 2021-01-20 02049, 2021

      • BrainzBot
        MBS-11339: Move Edit Artist Credit from /artist to /artist-credit
      • 2021-01-20 02051, 2021

      • ruaok
        ok, storage boxes support HTTPS and FTP. what if we get rid of our FTP setup and simply expose things we put into the storage box? that would simplify dump creation for LB, for sure.
      • 2021-01-20 02055, 2021

      • reosarevok
        (the new pages aren't on beta yet, but they're merged)
      • 2021-01-20 02034, 2021

      • alastairp
        ruaok: is that public ftp or private? from that page it sounds like they have a transfer limit externally, but it's unlimited internally
      • 2021-01-20 02053, 2021

      • alastairp
        which means that it might make sense to proxy through an existing server, given that they have unlimited external traffic too?
      • 2021-01-20 02024, 2021

      • ruaok
        yes, the page is not very clear -- what does 10 concurrent connections me? internal or external?
      • 2021-01-20 02003, 2021

      • ruaok
        but yes, that makes sense overall. serve the FTP from williams via the storage box. and then any of our servers which need to publish dumps, can publish them to the storage box, not to williams.
      • 2021-01-20 02001, 2021

      • Gazooo79494 has quit
      • 2021-01-20 02049, 2021

      • Gazooo79494 joined the channel
      • 2021-01-20 02011, 2021

      • yvanzo
        reosarevok: it's your own ticket, of course it seems sensible to you :D
      • 2021-01-20 02004, 2021

      • yvanzo
        reosarevok: there is no redirect mechanism for artist credits and they do not have MBIDs either. I’m not sure how you would implement that. It probably requires a schema change.
      • 2021-01-20 02052, 2021

      • yvanzo
        reosarevok: maybe /artist/…/credit/… would be a better place than the currently merged /artist-credit/… so to clarify artist-credits depend on artist?
      • 2021-01-20 02003, 2021

      • antlarr2 joined the channel
      • 2021-01-20 02032, 2021

      • antlarr has quit
      • 2021-01-20 02040, 2021

      • reosarevok
        It doesn't though. They have their own IDs. We don't need a redirect mechanism, the original links already include the ac id
      • 2021-01-20 02041, 2021

      • Sophist-UK
      • 2021-01-20 02042, 2021

      • Sophist-UK
      • 2021-01-20 02001, 2021

      • Sophist-UK
      • 2021-01-20 02002, 2021

      • Sophist-UK
      • 2021-01-20 02003, 2021

      • Sophist-UK
      • 2021-01-20 02004, 2021

      • Sophist-UK
      • 2021-01-20 02005, 2021

      • Sophist-UK
      • 2021-01-20 02006, 2021

      • Sophist-UK
      • 2021-01-20 02007, 2021

      • Sophist-UK
      • 2021-01-20 02008, 2021

      • Sophist-UK
      • 2021-01-20 02009, 2021

      • Sophist-UK
      • 2021-01-20 02010, 2021

      • Sophist-UK
      • 2021-01-20 02011, 2021

      • Sophist-UK
      • 2021-01-20 02012, 2021

      • Sophist-UK
      • 2021-01-20 02024, 2021

      • Sophist-UK
        Oops - sorry.
      • 2021-01-20 02004, 2021

      • yvanzo
        reosarevok: I mean, having a top level URL like /artist-credit/… may let think that this is a permanent link like /artist/…; this is not the case because there is no redirect mechanism.
      • 2021-01-20 02037, 2021

      • yvanzo
        If only URLs such as /artist/…/credit/… are provided, it can at least be redirected to the related artist page.
      • 2021-01-20 02019, 2021

      • Etua has quit
      • 2021-01-20 02043, 2021

      • Sophist_UK joined the channel
      • 2021-01-20 02011, 2021

      • Sophist-UK has quit
      • 2021-01-20 02057, 2021

      • diru1100 has quit
      • 2021-01-20 02044, 2021

      • sumedh joined the channel
      • 2021-01-20 02037, 2021

      • loujine
        zas: I get random timeouts when querying test.musicbrainz.org/ws/2/ (sometimes answers in < 1 sec, sometimes > 10 sec)
      • 2021-01-20 02049, 2021

      • loujine
        no problem on mb.o and beta.mb.o
      • 2021-01-20 02009, 2021

      • loujine
        just so that you know, it's not a real problem for me
      • 2021-01-20 02016, 2021

      • sumedh has quit
      • 2021-01-20 02023, 2021

      • sumedh joined the channel
      • 2021-01-20 02038, 2021

      • sumedh has quit
      • 2021-01-20 02026, 2021

      • sumedh joined the channel
      • 2021-01-20 02013, 2021

      • ruaok
        pristine___: should we merge that massive PR and work on a release?
      • 2021-01-20 02055, 2021

      • pristine___
        Yes!
      • 2021-01-20 02055, 2021

      • ruaok
        ok, I need one other PR merged before that. let me chase alastairp for this...
      • 2021-01-20 02029, 2021

      • alastairp
        the prod release of typesense?
      • 2021-01-20 02042, 2021

      • ruaok
        sentry cleanup.
      • 2021-01-20 02048, 2021

      • alastairp
        ah
      • 2021-01-20 02002, 2021

      • ruaok
        let me merger master in to fix that failing test.
      • 2021-01-20 02021, 2021

      • alastairp
        I made one comment about the format of the title of the inactive user messages
      • 2021-01-20 02058, 2021

      • alastairp
        along the lines of saying if the aim is to reduce the number of emails that we get, making the title consistent is going to prevent sentry from emailing the alert for each user
      • 2021-01-20 02059, 2021

      • ruaok
        yep, but I see no reason to deal with that, if it isn't clear that we need to keep that message.
      • 2021-01-20 02019, 2021

      • alastairp
        will changing from error to debug stop sentry from sending emails?
      • 2021-01-20 02008, 2021

      • ruaok
        one way to deal with that.
      • 2021-01-20 02017, 2021

      • alastairp
        heh, that certainly is a fix
      • 2021-01-20 02025, 2021

      • ruaok
        right. now all 4 tests failed. is our jenkins config borked?
      • 2021-01-20 02023, 2021

      • pristine___
        Does the Jenkins integration test fail because of that, I mean the PR which you want to merge beforehand
      • 2021-01-20 02055, 2021

      • alastairp
        yes, it looks like docker-compose inside jenkins can't connect to the docker socket: https://ci.metabrainz.org/job/listenbrainz-integr…
      • 2021-01-20 02020, 2021

      • alastairp
        I'd restart the jenkins container, that should do it
      • 2021-01-20 02042, 2021

      • ruaok
        prior to the new issue that cropped up, there was a failing test that we have no idea why it is failing, so we suppressed it.
      • 2021-01-20 02052, 2021

      • alastairp
        I'm away until 5:30 with teaching, will be back after that to answer any questions if they come up
      • 2021-01-20 02007, 2021

      • alastairp
        there was the disk issue on williams this morning, which I think is probably the cause
      • 2021-01-20 02022, 2021

      • ruaok
        I'll deal with the two issues (one which is not of my doing) and merge and proceed to release.
      • 2021-01-20 02001, 2021

      • ruaok
        zas: ping
      • 2021-01-20 02029, 2021

      • zas
        pong
      • 2021-01-20 02047, 2021

      • ruaok
        hey.
      • 2021-01-20 02055, 2021

      • ruaok
        so, i have a bit of a quandry...
      • 2021-01-20 02013, 2021

      • ruaok
        typsense wants and API key defined as part of its command line arguments for invoking the service.
      • 2021-01-20 02031, 2021

      • ruaok
        its a service that requires no changes on our part, its just running an image from docker hub.
      • 2021-01-20 02037, 2021

      • yvanzo
        I can drop one dump of MB search indexes (30GB) if that helps.
      • 2021-01-20 02058, 2021

      • ruaok
        yvanzo: yes please, if that isn't a bother.
      • 2021-01-20 02018, 2021

      • ruaok
        zas: and I would like to keep the API in docker-server-configs.
      • 2021-01-20 02042, 2021

      • ruaok
        so the start_services.sh script needs to fetch the key from consul. or have the key be duplicated.
      • 2021-01-20 02056, 2021

      • ruaok
      • 2021-01-20 02057, 2021

      • yvanzo
        Will do, that will effective at very short term only as next dump is next saturday, and it will be larger: 45GB.
      • 2021-01-20 02007, 2021

      • ruaok
        see start_consul_agent line 9
      • 2021-01-20 02014, 2021

      • zas
        hmmm, is there another way to get pass the key?
      • 2021-01-20 02023, 2021

      • ruaok
        let me see
      • 2021-01-20 02010, 2021

      • zas
        because passing it through command line is rather unsecure (since everyone can see it)
      • 2021-01-20 02015, 2021

      • zas
        /etc/typesense/typesense-server.ini
      • 2021-01-20 02048, 2021

      • ruaok
        well, the key isn't super important. I want all of our infrastructure to be able to access it.
      • 2021-01-20 02017, 2021

      • zas
        so you can use docker-template to generate /etc/typesense/typesense-server.ini but if you use stock image it doesn't support consul-template
      • 2021-01-20 02053, 2021

      • ruaok
        I can mount the ini file via a volume...
      • 2021-01-20 02059, 2021

      • zas
        so the other approach: mount a volume -v typesense-server.ini:/etc/typesense/typesense-server.ini
      • 2021-01-20 02003, 2021

      • zas
        yes^^
      • 2021-01-20 02008, 2021

      • ruaok
        ok, I'll do that. :)
      • 2021-01-20 02044, 2021

      • ruaok
        Mr_Monkey: does this javascript test failure make much sense to you? https://ci.metabrainz.org/job/listenbrainz-js/172…
      • 2021-01-20 02013, 2021

      • ruaok
        zas: actually, we didn't answer my question that I had in the first place.
      • 2021-01-20 02054, 2021

      • ruaok
        fetching the API key, which should be stored in consul, as part of the start_services script -- do you see any problems with that?
      • 2021-01-20 02026, 2021

      • zas
        well, if you want to fetch it from consul you can simply use curl + consul http api
      • 2021-01-20 02049, 2021

      • ruaok
        yes, that was my plan, but do you see any problems with that approach?