=== Pici is now known as Guest55695
[20:05] <xnox> how often does lp.net update it's debian mirror?
[20:05] <xnox> $ requestsync clasp
[20:05] <xnox> W: Target release missing - assuming quantal
[20:05] <xnox> E: The version in Ubuntu (2.0.6-1ubuntu1) is newer than the version in Debian (2.0.6-1). Aborting.
[20:06] <Laney> can be up to a day ime
[20:07] <cjwatson> I'm told the importer runs every six hours, although I'm not sure that *entirely* tallies with my experience
[20:08] <cjwatson> 10 3,9,15,21 * * * /srv/debian-import.launchpad.net/production/launchpad/scripts/gina.py squeeze sid experimental wheezy -q --log-file=INFO:/srv/debian-import.launchpad.net/production-logs/gina.log
[20:08] <Laney> doesn't it pull from the UK mirror?
[20:08] <Laney> so there are several sources of delay in between
[20:08] <cjwatson> Slightly more complete:
[20:08] <cjwatson> # Mirror from ftp.uk.debian.org.  Debian publishes at "52 1,7,13,19"
[20:08] <cjwatson> # but we also need to allow time for the above mirror to update.
[20:08] <cjwatson> 05 3,9,15,21 * * * sh /srv/debian-import.launchpad.net/scripts/mirror-update.sh >> /srv/debian-import.launchpad.net/production-logs/mirror-update.log 2>&1
[20:08] <cjwatson> # Allow 5 minutes between mirroring and running Gina. The mirror logs suggest
[20:09] <cjwatson> # that mirroring takes at most about 1m30s.
[20:09] <cjwatson> 10 3,9,15,21 * * * /srv/debian-import.launchpad.net/production/launchpad/scripts/gina.py squeeze sid experimental wheezy -q --log-file=INFO:/srv/debian-import.launchpad.net/production-logs/gina.log
[20:09] <cjwatson> I don't know why those two jobs aren't a single job with &&; might ask about that
[20:09] <cjwatson> (but during the week)
[20:09] <cjwatson> Oh, mind you, .au might have started
[20:13] <xnox> ok, thanks. Back to drinking sunday evening tea =)
[20:13] <cjwatson> Hm, I think it might be optimistic about Debian's publishing time
[20:13] <cjwatson> So we might be hitting near worst-case latency
[20:13] <cjwatson> I'll keep an eye on it as best I can and have a look
[20:20] <cjwatson> xnox: This is basically bug 798611
[20:20] <ubot2> Launchpad bug 798611 in launchpad "Package Auto Sync seems to get ahead of +localdiff calculation" [High,Triaged] https://launchpad.net/bugs/798611
=== yofel_ is now known as yofel
[21:16] <tumbleweed> or if we really wanted to optimise it, organise push mirroring
[21:52] <Daviey> Does config changes for status.ubuntu.com get rolled out automagically now, or does it still require a human to pull?
[21:52]  * xnox is this sync running....
[21:56] <cjwatson> xnox: the new binaries?  yes, I run syncs more or less daily
[21:57] <cjwatson> you can ignore queuebot if it's not interesting :)
[21:58] <xnox> cjwatson: ok. I'm just pondering when will I finally be able to requestsync clasp.
[21:59] <xnox> I am doing ubuntu development DD styly: "TeamUpload/NMU + sync to ubuntu" =))))
[21:59] <xnox> s/styly/style/
[21:59] <cjwatson> my investigations earlier suggested that the maximum latency is ~13 hours.  but remember that that's from Debian actually publishing to mirrors, not from the upload
[22:00] <cjwatson> the clasp source is still in incoming.debian.org
[22:00] <cjwatson> oh, it's on ftp as well
[22:01] <cjwatson> a little surprising it's not on LP now
[22:01] <Laney> push could be an idea
[22:01] <cjwatson> but given it's not, I'd expect it to arrive in the import that finishes around 03:40
[22:01] <cjwatson> (UTC)
[22:02] <cjwatson> unless the import starting at 21:10 UTC is slower than I've been led to believe, I guess
[22:02] <cjwatson> I don't have easy access to the logs
[22:03] <cjwatson> moving the import back an hour is probably an order of magnitude less effort than push mirroring, even if it's hacky :)
[22:03] <Daviey> If you want it sooner, uploading with ~ubuntu1, then syncing out the package when LP knows about it is the best of all evils IMO.
[22:03] <cjwatson> I doubt clasp is that urgent
[22:03] <xnox> Daviey: if I had upload rights in ubuntu....
[22:04] <Daviey> Encourages push to Debian first, avoids having an essential fakesync (different sigs).. and still a lower version allowing a sync to override
[22:04] <cjwatson> I'd really prefer people not do that unless they have a damn good reason please
[22:04] <cjwatson> uses buildd resources for minor benefit
[22:04] <cjwatson> and it's easy to screw it up and end up requiring a fakesync in future
[22:04] <Daviey> cjwatson: right, 'if it's urgent'
[22:04] <cjwatson> *very urgent
[22:05] <cjwatson> anyway, it looks like we can easily make the importer have considerably less latency, and I'll look into that tomorrow
[22:05] <xnox> awesome =)