#metabrainz

/

      • reosarevok
        I did not!
      • 2017-12-20 35424, 2017

      • reosarevok
        Here at least people with fire stoves can just burn that stuff, but there...
      • 2017-12-20 35438, 2017

      • ruaok
        they came up with something creative. let me see if I can find it.
      • 2017-12-20 35446, 2017

      • ruaok
      • 2017-12-20 35452, 2017

      • ruaok doesn't share the disgusting video with everyone else
      • 2017-12-20 35413, 2017

      • iliekcomputers
        alastairp: I've made changes to gh:AB#233 and gh:AB#222. I think they should be ready for merge now, please take a look when you get the time :)
      • 2017-12-20 35435, 2017

      • iliekcomputers
        hmm, seems like BrainzBot doesn't link AB PRs :?
      • 2017-12-20 35440, 2017

      • iliekcomputers
        gh:LB#201
      • 2017-12-20 35441, 2017

      • BrainzBot
        LB-177: Make influx-writer account for a range of time when looking for duplicates: https://github.com/metabrainz/listenbrainz-server…
      • 2017-12-20 35447, 2017

      • iliekcomputers
        gh:AB#222
      • 2017-12-20 35409, 2017

      • iliekcomputers
        weird
      • 2017-12-20 35419, 2017

      • ruaok
        gah, I haven't looked at AB in such a long time. :(
      • 2017-12-20 35448, 2017

      • samj1912
        iliekcomputers: you need to add the alias
      • 2017-12-20 35406, 2017

      • samj1912
        BrainzBot: gh:AB=acousticbrainz-server
      • 2017-12-20 35406, 2017

      • BrainzBot
        Successfully stored the repo acousticbrainz-server as AB for Github lookups
      • 2017-12-20 35414, 2017

      • samj1912
        gh:AB#222
      • 2017-12-20 35415, 2017

      • BrainzBot
        AB-138 Allow pasting URLs for recording MBIDs: https://github.com/metabrainz/acousticbrainz-serv…
      • 2017-12-20 35418, 2017

      • iliekcomputers
        ruaok: we've made some progress with AB in the past few days, going through the PR backlog :)
      • 2017-12-20 35421, 2017

      • iliekcomputers
        samj1912: thanks!
      • 2017-12-20 35440, 2017

      • samj1912
        o/
      • 2017-12-20 35401, 2017

      • drsaunders has quit
      • 2017-12-20 35433, 2017

      • drsaunders joined the channel
      • 2017-12-20 35446, 2017

      • naiveai joined the channel
      • 2017-12-20 35457, 2017

      • sampsyo has quit
      • 2017-12-20 35400, 2017

      • jesus2099 joined the channel
      • 2017-12-20 35430, 2017

      • sampsyo joined the channel
      • 2017-12-20 35431, 2017

      • samj1912
      • 2017-12-20 35439, 2017

      • samj1912
        It is finally working (fingers crossed)
      • 2017-12-20 35448, 2017

      • samj1912
        and no its not ending up in the retry queue :P
      • 2017-12-20 35413, 2017

      • samj1912
        ruaok: ^
      • 2017-12-20 35425, 2017

      • samj1912
        and I also figured out more on how to solve our throughput problem
      • 2017-12-20 35402, 2017

      • ruaok
        nice! how is that?
      • 2017-12-20 35415, 2017

      • samj1912
        currently sir is processing 1 days worth of backlog in about 8 minutes
      • 2017-12-20 35432, 2017

      • samj1912
        anyways, I was going through old sir PRs, looking at mineo and gentlecat's comments
      • 2017-12-20 35443, 2017

      • samj1912
        apparently sir was made to be run on multiple nodes
      • 2017-12-20 35457, 2017

      • samj1912
        and it plays nicely with rmq
      • 2017-12-20 35418, 2017

      • samj1912
      • 2017-12-20 35418, 2017

      • samj1912
        current rmq utilization is about 1%
      • 2017-12-20 35438, 2017

      • samj1912
        it can handle multiple consumers
      • 2017-12-20 35456, 2017

      • ruaok
        awesome.
      • 2017-12-20 35405, 2017

      • samj1912
        so if we have sir running on 2 or 3 nodes, it can easily work through anomalies
      • 2017-12-20 35408, 2017

      • ruaok
        so in theory we can run one consumer normally and monitor it.
      • 2017-12-20 35412, 2017

      • samj1912
        yup
      • 2017-12-20 35415, 2017

      • ruaok
        should it fall behind, we start more consumers.
      • 2017-12-20 35428, 2017

      • samj1912
        currently I put a limit on indexed entities
      • 2017-12-20 35439, 2017

      • samj1912
        if they go past a certain number they end up on a different queue
      • 2017-12-20 35451, 2017

      • samj1912
        this way we can have 1 sir instance just churning through large reindexes
      • 2017-12-20 35410, 2017

      • samj1912
      • 2017-12-20 35428, 2017

      • samj1912
        processing 30k entities, it encountered 33 cases where number was greater than 100k (configurable)
      • 2017-12-20 35430, 2017

      • alastairp
        iliekcomputers: gah, I'm in the middle of that
      • 2017-12-20 35434, 2017

      • alastairp
        OK, I'll merge your changes
      • 2017-12-20 35443, 2017

      • samj1912
        so we can have one sir instance doing timely updates
      • 2017-12-20 35457, 2017

      • samj1912
        and one instance taking care of things that will be blocking the normal one
      • 2017-12-20 35407, 2017

      • samj1912
        does that sound like a good plan?
      • 2017-12-20 35427, 2017

      • iliekcomputers
        sorry! they were small changes, so I thought I might as well...
      • 2017-12-20 35429, 2017

      • samj1912
      • 2017-12-20 35437, 2017

      • alastairp
        yeah, no worries
      • 2017-12-20 35443, 2017

      • samj1912
      • 2017-12-20 35455, 2017

      • alastairp
        I'm just making some other changes to 222 too, adding in "automatic submit" if the item is valid
      • 2017-12-20 35459, 2017

      • alastairp
        which is pretty cool
      • 2017-12-20 35408, 2017

      • samj1912
        so I started sir at 19:44, out of all the entities only 73 ended up in the retry queue (which was already filled upto 57 before I started the latest experiment)
      • 2017-12-20 35404, 2017

      • jesus2099 has left the channel
      • 2017-12-20 35458, 2017

      • ruaok
        samj1912: it sounds very reasonable.
      • 2017-12-20 35426, 2017

      • ruaok
        are the solr indexes being kept up to date now?
      • 2017-12-20 35435, 2017

      • iliekcomputers
        alastairp: 👍🏽
      • 2017-12-20 35436, 2017

      • samj1912
        and I am also logging everything with tracebacks so we can actually examine the queue for messages and fix bugs
      • 2017-12-20 35444, 2017

      • samj1912
        I figured out 2 more bugs right now
      • 2017-12-20 35457, 2017

      • samj1912
        ruaok: they can be now
      • 2017-12-20 35424, 2017

      • ruaok
        ok, when you fix those two bugs, keep the indexes updating.
      • 2017-12-20 35425, 2017

      • samj1912
        I will have to run a reindex first since I loaded the triggers 3 days back
      • 2017-12-20 35430, 2017

      • ruaok
        watch for a day or so.
      • 2017-12-20 35432, 2017

      • samj1912
        cool
      • 2017-12-20 35442, 2017

      • ruaok
        then we should switch beta over to use the new search.
      • 2017-12-20 35450, 2017

      • ruaok
        and then put out a massive call for testing.
      • 2017-12-20 35454, 2017

      • samj1912
        there are some bugs on config reloads that I need to iron out
      • 2017-12-20 35415, 2017

      • samj1912
        which I will do once I have indexing up and running perfectly
      • 2017-12-20 35413, 2017

      • ruaok
        great.
      • 2017-12-20 35426, 2017

      • samj1912
        also, I am still running at a modest config, so we can probably churn more out of sir
      • 2017-12-20 35431, 2017

      • ruaok
        remember that I promised you a $500 bonus if this was up and running before the end of the year?
      • 2017-12-20 35435, 2017

      • UmkaDK joined the channel
      • 2017-12-20 35442, 2017

      • samj1912
        yeah :P
      • 2017-12-20 35445, 2017

      • ruaok
        I think you've already earned that.
      • 2017-12-20 35453, 2017

      • ruaok
        so, make sure to add that to your jan invoice.
      • 2017-12-20 35456, 2017

      • UmkaDK_ has quit
      • 2017-12-20 35404, 2017

      • samj1912
        cool :D! thanks!
      • 2017-12-20 35408, 2017

      • ruaok
        I'm sooooo happy that you're rocking this. we've needed a hero like you!
      • 2017-12-20 35413, 2017

      • ruaok
        :D
      • 2017-12-20 35434, 2017

      • ruaok
        still, I worry about you and iliekcomputers. how many hours of sleep do you normally get?
      • 2017-12-20 35435, 2017

      • samj1912
        this feels so great after the last week or 2 of struggling like shit with it :P
      • 2017-12-20 35455, 2017

      • ruaok
        its been a long slog, not just the last two weeks. :)
      • 2017-12-20 35422, 2017

      • samj1912
      • 2017-12-20 35433, 2017

      • ruaok
        \ø/
      • 2017-12-20 35437, 2017

      • samj1912
      • 2017-12-20 35444, 2017

      • samj1912
      • 2017-12-20 35401, 2017

      • alastairp
        what does failed mean? temporary or permanent?
      • 2017-12-20 35424, 2017

      • samj1912
        alastairp: I currently have those for messages which end up querying more than 100k rows
      • 2017-12-20 35440, 2017

      • samj1912
        I had a bug earlier in sql generation which is why it hit 33
      • 2017-12-20 35452, 2017

      • alastairp
        oh right, so it's not exactly "was unable to index entity x"
      • 2017-12-20 35455, 2017

      • samj1912
        but I solved it, and in the latest churning through 50k messages not 1 hit 100k rows
      • 2017-12-20 35401, 2017

      • samj1912
        yup
      • 2017-12-20 35406, 2017

      • alastairp
        nice
      • 2017-12-20 35420, 2017

      • samj1912
        retry is caused on exceptions
      • 2017-12-20 35451, 2017

      • samj1912
        50k messages caused 21 messages to end up on the retry queue
      • 2017-12-20 35403, 2017

      • samj1912
        all of them correspond to the same bug
      • 2017-12-20 35429, 2017

      • samj1912
        and it took 15 minutes to process 50k messages
      • 2017-12-20 35413, 2017

      • samj1912
        I will push the latest changes to my PR now, this was a lot of test, debug, retry, repeat
      • 2017-12-20 35415, 2017

      • naiveai
        !m samj1912
      • 2017-12-20 35415, 2017

      • BrainzBot
        You're doing good work, samj1912!
      • 2017-12-20 35428, 2017

      • naiveai
        I have no idea what you're doing, but the charts look very fancy.
      • 2017-12-20 35433, 2017

      • naiveai
        :D
      • 2017-12-20 35446, 2017

      • reosarevok
        zas: prod server issues?
      • 2017-12-20 35409, 2017

      • samj1912
        nothing on alrets?
      • 2017-12-20 35413, 2017

      • samj1912
        let me check
      • 2017-12-20 35429, 2017

      • zas
        nothing i'm aware of, which kind of issues are you experiencing ?
      • 2017-12-20 35447, 2017

      • ruaok
        slow web pages.
      • 2017-12-20 35432, 2017

      • samj1912
        might be because of sir, but I kept the queries very limited
      • 2017-12-20 35400, 2017

      • samj1912
        let me check
      • 2017-12-20 35401, 2017

      • ruaok
        sir points to bowie or queen?
      • 2017-12-20 35413, 2017

      • zas
        one backend is slow, on hip, i'll check
      • 2017-12-20 35424, 2017

      • samj1912
        bowie, it was a disaster last time we had it point to queen, bitmap told me to use master
      • 2017-12-20 35452, 2017

      • samj1912
        sir runs on queen though
      • 2017-12-20 35457, 2017

      • samj1912
        we should move it elsewhere
      • 2017-12-20 35405, 2017

      • samj1912
        it takes a lot of cpu %
      • 2017-12-20 35422, 2017

      • Slurpee joined the channel
      • 2017-12-20 35422, 2017

      • Slurpee has quit
      • 2017-12-20 35422, 2017

      • Slurpee joined the channel
      • 2017-12-20 35428, 2017

      • naiveai
        haruute[m]: what're you working on rn?
      • 2017-12-20 35450, 2017

      • samj1912
      • 2017-12-20 35400, 2017

      • ruaok
        cpu on postgres master didn't seem to care about your efforts.
      • 2017-12-20 35418, 2017

      • samj1912
        sir doesn't seem to have affect pg-master much?
      • 2017-12-20 35429, 2017

      • ruaok
        I don't see the impact, do you?
      • 2017-12-20 35438, 2017

      • samj1912
        nope
      • 2017-12-20 35440, 2017

      • samj1912
        not at all
      • 2017-12-20 35456, 2017

      • samj1912
        phew
      • 2017-12-20 35411, 2017

      • haruute[m]
        naiveai: for now, i'am claiming anison generation url task
      • 2017-12-20 35417, 2017

      • samj1912
        I was keeping my fingers crossed :X
      • 2017-12-20 35426, 2017

      • naiveai
        haruute[m]: aight. will do smth else then
      • 2017-12-20 35439, 2017

      • samj1912
        ruaok: once sir is up and running, next we will need to take a look at solr
      • 2017-12-20 35446, 2017

      • Slurpee has quit
      • 2017-12-20 35450, 2017

      • samj1912
        I did a stress test on solr a month back
      • 2017-12-20 35428, 2017

      • samj1912
        a single node could handle about 30 req/s without much problems
      • 2017-12-20 35437, 2017

      • dragonzeron joined the channel
      • 2017-12-20 35443, 2017

      • samj1912
        our current search servers are hit 100reqs/s over 2 nodes
      • 2017-12-20 35409, 2017

      • ruaok gets a little excited.
      • 2017-12-20 35438, 2017

      • ruaok
        when we get this deployed this will be a project 10 years in the making, finally done.
      • 2017-12-20 35444, 2017

      • samj1912
        but we will have to probably end up using solr cloud
      • 2017-12-20 35448, 2017

      • samj1912
        yeah :D
      • 2017-12-20 35402, 2017

      • ruaok
        why do you say that?
      • 2017-12-20 35436, 2017

      • samj1912
        because solr shat itself as I started bombarding it with 45 req/s (while I was doing a reindex of recordings alongside)
      • 2017-12-20 35405, 2017

      • samj1912
        and I was only querying the recording core