#metabrainz

/

      • ruaok
        moooin! \ø/
      • 2018-06-01 15234, 2018

      • ruaok hopes that rajoy gets tossed into the gutter
      • 2018-06-01 15241, 2018

      • ruaok
        may his government follow
      • 2018-06-01 15220, 2018

      • samj1912
        elections?
      • 2018-06-01 15251, 2018

      • samj1912
        ah no, scandal :P
      • 2018-06-01 15229, 2018

      • ruaok
        neverending scandals.
      • 2018-06-01 15240, 2018

      • ruaok
        if he gets tossed, I would expect a massive street party tonight.
      • 2018-06-01 15245, 2018

      • samj1912
        lol
      • 2018-06-01 15205, 2018

      • UmkaDK joined the channel
      • 2018-06-01 15230, 2018

      • drsaunder has quit
      • 2018-06-01 15243, 2018

      • drsaunder joined the channel
      • 2018-06-01 15255, 2018

      • antlarr has quit
      • 2018-06-01 15233, 2018

      • antlarr joined the channel
      • 2018-06-01 15240, 2018

      • KassOtsimine has quit
      • 2018-06-01 15204, 2018

      • KassOtsimine joined the channel
      • 2018-06-01 15202, 2018

      • Freso
        GeneralDiscourse: Are you using the VM?
      • 2018-06-01 15246, 2018

      • Freso
        Though I guess we haven't had a schema change since the VM was made.
      • 2018-06-01 15246, 2018

      • Freso
        GeneralDiscourse: bitmap or yvanzo should be able to help you, whenever one of them is around (and you are too). You can also perhaps ask on our forums if you can't stick around on IRC (but I guess you're still here almost 8 hours later, so… :)): https://community.metabrainz.org/
      • 2018-06-01 15238, 2018

      • samj1912
        zas, ping me when you are free
      • 2018-06-01 15258, 2018

      • samj1912
        Let's look into the next step for cloud deployment
      • 2018-06-01 15219, 2018

      • outsidecontext joined the channel
      • 2018-06-01 15234, 2018

      • zas
        samj1912: i am around
      • 2018-06-01 15240, 2018

      • samj1912
        Cool
      • 2018-06-01 15247, 2018

      • samj1912
        What's next then?
      • 2018-06-01 15237, 2018

      • HSOWA joined the channel
      • 2018-06-01 15237, 2018

      • HSOWA has quit
      • 2018-06-01 15237, 2018

      • HSOWA joined the channel
      • 2018-06-01 15254, 2018

      • zas
        first, we need to decide about shards & replicas, the current config has no replica
      • 2018-06-01 15243, 2018

      • zas
        samj1912: did you document how you loaded config in zookeeper etc ?
      • 2018-06-01 15222, 2018

      • reosarevok
        ruaok: no big party probs because the alternative isn't thaaaat much better
      • 2018-06-01 15232, 2018

      • reosarevok
        But still, small victories
      • 2018-06-01 15206, 2018

      • samj1912
        zas: yup
      • 2018-06-01 15222, 2018

      • ruaok
        We'll see. There is plenty up hate for him here.
      • 2018-06-01 15223, 2018

      • samj1912
        let me push the changes to the repo
      • 2018-06-01 15236, 2018

      • ruaok
        42 for, 25 against. :-)
      • 2018-06-01 15247, 2018

      • ruaok
        .. pent up ...
      • 2018-06-01 15255, 2018

      • KassOtsimine has quit
      • 2018-06-01 15209, 2018

      • Freso
        samj1912: https://community.metabrainz.org/t/isrcs-json-ret… seems similar to what the issue was with the SOLR recording search pages ISRC'ing, no? Not sure if it's directly related though. Maybe yvanzo or bitmap could take a look.
      • 2018-06-01 15218, 2018

      • Freso
        XD
      • 2018-06-01 15224, 2018

      • Freso
        s/ISRC'ing/ISE'ing/
      • 2018-06-01 15246, 2018

      • outsidecontext has quit
      • 2018-06-01 15219, 2018

      • samj1912
        Freso: yup
      • 2018-06-01 15222, 2018

      • zas
        samj1912: on mb-solr-2; i connect to zk using /opt/zookeeper/bin/zkCli.sh
      • 2018-06-01 15233, 2018

      • samj1912
        I guess he was testing on test.mb
      • 2018-06-01 15237, 2018

      • zas
      • 2018-06-01 15228, 2018

      • zas
        it should list collections, i wonder if this output is correct
      • 2018-06-01 15205, 2018

      • zas
        oh, perhaps not
      • 2018-06-01 15208, 2018

      • zas
      • 2018-06-01 15224, 2018

      • zas
        says "clusterstate.json is a remnant of when all collection information was held in that node. It should always be empty currently."
      • 2018-06-01 15215, 2018

      • samj1912
      • 2018-06-01 15216, 2018

      • samj1912
        zas:
      • 2018-06-01 15208, 2018

      • samj1912
        I guess I'll also need to add instructions to add the query response writer
      • 2018-06-01 15217, 2018

      • Freso
        chhavi_: I think this is of interest to you: https://tickets.metabrainz.org/browse/MBS-9732 ?
      • 2018-06-01 15218, 2018

      • BrainzBot
        MBS-9732: Change "expires in" wording/phrasing
      • 2018-06-01 15224, 2018

      • samj1912
      • 2018-06-01 15252, 2018

      • samj1912
        my bash scripting game isn't upto the mark :P
      • 2018-06-01 15205, 2018

      • samj1912
        so please feel free to write the for loops :P
      • 2018-06-01 15247, 2018

      • samj1912
        You can use both zkcli or the solr CLI
      • 2018-06-01 15252, 2018

      • samj1912
        I prefer the solr cli
      • 2018-06-01 15212, 2018

      • samj1912
      • 2018-06-01 15258, 2018

      • zas
        about point 2, copying the repo is a bad idea, since it will copy .git too, instead you should use git archive
      • 2018-06-01 15209, 2018

      • reosarevok
        ruaok: whee
      • 2018-06-01 15217, 2018

      • zas
        something like TARGET=/tmp/config; mkdir -p $TARGET; git archive --format tar HEAD |tar xvf - -C $TARGET
      • 2018-06-01 15218, 2018

      • reosarevok
        I guess they can't fuck it up more :)
      • 2018-06-01 15228, 2018

      • UmkaDK_ joined the channel
      • 2018-06-01 15233, 2018

      • zas
        what a pain it doesn't support symlinks...
      • 2018-06-01 15242, 2018

      • ruaok
        YES YES YES YES YES!!!!
      • 2018-06-01 15243, 2018

      • samj1912
        yup :P
      • 2018-06-01 15227, 2018

      • UmkaDK has quit
      • 2018-06-01 15212, 2018

      • samj1912
        zas: yeah, makes sense
      • 2018-06-01 15247, 2018

      • zas
        samj1912: edit the README directly no need to use PRs
      • 2018-06-01 15223, 2018

      • samj1912
        cool
      • 2018-06-01 15243, 2018

      • ruaok
        samj1912: it says money should arrive in 7 minutes. set a timer and look!
      • 2018-06-01 15246, 2018

      • ruaok doubt it
      • 2018-06-01 15247, 2018

      • zas
        Please add exemples for tools you are using (solr cli)
      • 2018-06-01 15249, 2018

      • samj1912
        woah :O
      • 2018-06-01 15229, 2018

      • ruaok
        and in theory you should receive the full, exact invoice amount.
      • 2018-06-01 15259, 2018

      • samj1912
        ruaok: woah O.o
      • 2018-06-01 15203, 2018

      • samj1912
        its here, and exact
      • 2018-06-01 15209, 2018

      • zas
        samj1912: about shards & replicas, do you have a clear idea (i don't) ? It seems to me we need autosharding and at least 2 replicas per collection
      • 2018-06-01 15218, 2018

      • ruaok
        in 5 minutes?
      • 2018-06-01 15219, 2018

      • ruaok
        holy shit.
      • 2018-06-01 15237, 2018

      • reosarevok
        :)
      • 2018-06-01 15222, 2018

      • samj1912
        zas: sorta, autosharding yes, replicas will depend on our needs
      • 2018-06-01 15242, 2018

      • zas
        ruaok: i used a hammer to crush applebot yesterday, it was making tons of requests very fast, apparently ignoring robots.txt, i blacklisted the IP block (faster than filtering on UA), but it was to restore decent response times for our users.
      • 2018-06-01 15206, 2018

      • ruaok
        transferwise is more expensive for a regular US bank transfer. hmm.
      • 2018-06-01 15220, 2018

      • ruaok
        zas: great, keep it like that.
      • 2018-06-01 15223, 2018

      • zas
      • 2018-06-01 15222, 2018

      • zas
      • 2018-06-01 15235, 2018

      • ruaok
        huh, again our bank is cheaper than transferwise.
      • 2018-06-01 15252, 2018

      • ruaok
        in the end I may only use it to pay our INR contractors.
      • 2018-06-01 15213, 2018

      • ruaok
        except zas and yvanzo, who get shafted with fees.
      • 2018-06-01 15214, 2018

      • samj1912
        ruaok: what's the inr fees?
      • 2018-06-01 15219, 2018

      • Darkloke joined the channel
      • 2018-06-01 15238, 2018

      • ruaok
        7.30 EUR
      • 2018-06-01 15215, 2018

      • samj1912
        almost 1.5 times the fee that is cut generally
      • 2018-06-01 15235, 2018

      • ruaok
        from your end.
      • 2018-06-01 15240, 2018

      • ruaok
        half from normal on our end.
      • 2018-06-01 15251, 2018

      • samj1912
        ooh
      • 2018-06-01 15255, 2018

      • ruaok
        so, transferwise is winning on the INR front.
      • 2018-06-01 15209, 2018

      • ruaok
        and it will win for france as well and possibly sweden.
      • 2018-06-01 15250, 2018

      • samj1912
        nice!
      • 2018-06-01 15219, 2018

      • ruaok runs off to get divorced
      • 2018-06-01 15241, 2018

      • samj1912
        zas: what's the current search req/s?
      • 2018-06-01 15258, 2018

      • samj1912
        zas: I guess we will also need to enable https://lucene.apache.org/solr/guide/6_6/distribu… ?
      • 2018-06-01 15221, 2018

      • rsh7 joined the channel
      • 2018-06-01 15209, 2018

      • zas
      • 2018-06-01 15203, 2018

      • zas
        mean peak at ~130 q/s (hitting upstreams), low around ~80
      • 2018-06-01 15238, 2018

      • zas
        for all 3 servers
      • 2018-06-01 15217, 2018

      • zas
        so to be safe we need at least 2 times the max, around 250 q/s, and it's a min
      • 2018-06-01 15251, 2018

      • KassOtsimine joined the channel
      • 2018-06-01 15252, 2018

      • zas
        on 3 nodes that's around 80q/s, 3 times what we get now
      • 2018-06-01 15205, 2018

      • samj1912
        Hmm
      • 2018-06-01 15214, 2018

      • zas
        but i'm pretty sure the current setup is under performing
      • 2018-06-01 15259, 2018

      • zas
        what is the current caching config ?
      • 2018-06-01 15215, 2018

      • samj1912
        zas, for load balancing have you taken a look at https://lucene.apache.org/solr/6_6_0//solr-solrj/…
      • 2018-06-01 15224, 2018

      • samj1912
        zas, not sure
      • 2018-06-01 15237, 2018

      • samj1912
        I will have to check
      • 2018-06-01 15209, 2018

      • zas
        samj1912: that's not for our case, that's for client lib
      • 2018-06-01 15223, 2018

      • HSOWA has quit
      • 2018-06-01 15226, 2018

      • zas
        we need a frontend like nginx reverse proxy or haproxy
      • 2018-06-01 15246, 2018

      • zas
        but we don't need to set it up yet, let's focus on the performance of one node first
      • 2018-06-01 15249, 2018

      • samj1912
        Oh cool
      • 2018-06-01 15209, 2018

      • samj1912
        Can you conduct your own benchmarks once?
      • 2018-06-01 15219, 2018

      • samj1912
        I am not sure if I did it correctly?
      • 2018-06-01 15242, 2018

      • zas
        can you document how you did ? i'll use it as a basis
      • 2018-06-01 15209, 2018

      • zas
        my feeling is that the current configuration is totally "untuned"
      • 2018-06-01 15253, 2018

      • samj1912
        I just varied a couple of params for ab, concurrency, the number of rows returned(25,100,300), and the output type (m json vs mbxml)
      • 2018-06-01 15212, 2018

      • Darkloke
        Hi2All. What is this type of error in picard - "E: 13:39:19 Fingerprint calculator failed exit code = 2, exit status = 0, error = Unknown error" ?
      • 2018-06-01 15236, 2018

      • zas
        how did you generate queries ? can you show url schema ?
      • 2018-06-01 15243, 2018

      • samj1912
        I ran the default query under the advanced end point
      • 2018-06-01 15205, 2018

      • samj1912
      • 2018-06-01 15205, 2018

      • zas
        did you limit number of results or anything?
      • 2018-06-01 15212, 2018

      • samj1912
        See above
      • 2018-06-01 15226, 2018

      • samj1912
        As, I said, varied between 25, 100 and 300
      • 2018-06-01 15240, 2018

      • zas
        but this query returns all releases ?
      • 2018-06-01 15252, 2018

      • samj1912
        Yup
      • 2018-06-01 15221, 2018

      • samj1912
        But shows the top 30 results
      • 2018-06-01 15258, 2018

      • zas
        do you want me to extract a bunch of real queries from current logs ? and we use same params to mimic actual traffic
      • 2018-06-01 15216, 2018

      • samj1912
        Sure
      • 2018-06-01 15225, 2018

      • samj1912
        That will work better
      • 2018-06-01 15253, 2018

      • samj1912
        But we will have to alter the params a bit
      • 2018-06-01 15211, 2018

      • samj1912
        Solr has different params than ws
      • 2018-06-01 15241, 2018

      • zas
        yup, i'll first extract requests
      • 2018-06-01 15232, 2018

      • zas
        let's work on releases and recordings ok ?
      • 2018-06-01 15241, 2018

      • samj1912
        Okay
      • 2018-06-01 15254, 2018

      • samj1912
        Btw, recordings are still being indexed
      • 2018-06-01 15228, 2018

      • samj1912
        Or not
      • 2018-06-01 15239, 2018

      • samj1912
        Looks like it died again
      • 2018-06-01 15242, 2018

      • samj1912
        Sigh