Last modified: 2014-02-25 12:26:49 UTC
Although new tsv files typically appear some time after 08:00 on stat1002, today's (2014-02-06) sampled-1000 tsv file is not yet there [1] some 4 hours afterwards. Due to the file not being there, daily jobs that rely on the daily files being in place break. For example, today's wikipedia-zero run just failed. [1] ___________________________________________________________ qchris@stat1002 // 0 // 12:08:05 cwd: ~ ll /a/squid/archive/sampled/sampled-1000.tsv.log-201402* -rw-r--r-- 1 stats stats 646898016 Feb 1 06:25 /a/squid/archive/sampled/sampled-1000.tsv.log-20140201.gz -rw-r--r-- 1 stats stats 646896549 Feb 2 06:25 /a/squid/archive/sampled/sampled-1000.tsv.log-20140202.gz -rw-r--r-- 1 stats stats 739443262 Feb 3 06:25 /a/squid/archive/sampled/sampled-1000.tsv.log-20140203.gz -rw-r--r-- 1 stats stats 728781897 Feb 4 06:25 /a/squid/archive/sampled/sampled-1000.tsv.log-20140204.gz -rw-r--r-- 1 stats stats 723531018 Feb 5 06:25 /a/squid/archive/sampled/sampled-1000.tsv.log-20140205.gz
Checking on emery shows that the sampled-1000 file is ready on emery. Let's see if it syncs over toworrow.
Prioritization and scheduling of this bug is tracked on Mingle card https://wikimedia.mingle.thoughtworks.com/projects/analytics/cards/cards/1432
Going through recent changes in the puppet repo, it seems the merge of e405e7e622cb1a275b3c2acb65aad0ee4c1d7729 in the puppet repo is related: https://gerrit.wikimedia.org/r/#/c/110382
Nuria has made some changes in the mingle ticket to fix this; will confirm tomorrow that issue is resolved.
Change 112233 had a related patch set uploaded by QChris: Mark 2014-02-05, and 2014-02-06 as bad dates https://gerrit.wikimedia.org/r/112233
Verified that there are sampled logs for the last couple of days: 719726187 Feb 9 06:25 sampled-1000.tsv.log-20140209.gz 811851138 Feb 10 06:25 sampled-1000.tsv.log-20140210.gz Bug can be closed.
Good to see the new files on stat1002 again! Thanks. But since this was no outtake but a filter change, we should have good tsvs lying around on emery for stat1002's missing/bad files. Could you bring the good files over so we can rerun the jobs that failed due to the missing/bad tsvs? Also, could you please update the corresponding row in documentation table on https://wikitech.wikimedia.org/wiki/Analytics/Requests_stream ?
I will put in an R2 request to get access to emery
Sorry, "RT" request.
Missing file has been restored.
(In reply to nuria from comment #10) > Missing file has been restored. No more missing tsvs. Thanks! But comment 7 mentions two other things, which did not yet happen. Hence, reopening. (In reply to christian from comment #7) > [ bad files ] Just doing a plain “ll” on stat1002, the file for 2014-02-07 looks wrong. ___________________________________________________________ qchris@stat1002 // 0 // 10:43:43 cwd: ~ cd /a/squid/archive/sampled ; ll sampled-1000.tsv.log-2014020* -rw-r--r-- 1 stats stats 646898016 Feb 1 06:25 sampled-1000.tsv.log-20140201.gz -rw-r--r-- 1 stats stats 646896549 Feb 2 06:25 sampled-1000.tsv.log-20140202.gz -rw-r--r-- 1 stats stats 739443262 Feb 3 06:25 sampled-1000.tsv.log-20140203.gz -rw-r--r-- 1 stats stats 728781897 Feb 4 06:25 sampled-1000.tsv.log-20140204.gz -rw-r--r-- 1 stats stats 723531018 Feb 5 06:25 sampled-1000.tsv.log-20140205.gz -rw-r--r-- 1 stats stats 728198594 Feb 6 06:25 sampled-1000.tsv.log-20140206.gz -rw-r--r-- 1 stats stats 1022924002 Feb 7 06:25 sampled-1000.tsv.log-20140207.gz -rw-r--r-- 1 stats stats 753216484 Feb 8 06:25 sampled-1000.tsv.log-20140208.gz -rw-r--r-- 1 stats stats 719726187 Feb 9 06:25 sampled-1000.tsv.log-20140209.gz (In reply to christian from comment #7) > Also, could you please update the corresponding row in > documentation table on > https://wikitech.wikimedia.org/wiki/Analytics/Requests_stream > ? This has not happened either.
Nuria pinged me that the 20140207 file is available on stat1002. Thanks for the file! The new file seems to miss a few lines, but since it is a sampled file anyway, and the drop is way below the usual network drop, it's fine by me.
The good file for 20140207 is gone again, and we're back with the ~1GB file for 20140207 :-( @Nuria, are you sure you did copy the fixed file to the rsync source as we said yesterday in IRC: Feb 20 12:25:28 <qchris> Afterwards get it to the rsync source. ?
Sorry, I missunderstood rsync source, I thought you were talking about the directory in stat1002. It is fixed now so file should not change going forward.
(In reply to nuria from comment #14) > Sorry, I missunderstood rsync source, [...] No worries. The file is there again \o/ Thanks! The file owner changed, but meh. It is world readable anyways :-) But could you please clean up the left behind cruft file, as that would sooner or later get in the way when doing adhoc checks?
All deleted and good.
@Nuria: The files now look good. Thanks \o/ But again ... could you please also update the corresponding row in documentation table on https://wikitech.wikimedia.org/wiki/Analytics/Requests_stream ?
Documentation has been updated by Nuria.