| lool | wgrant: hey, I'm trying to do a livefs build against a private PPA, but it seems to be stuck after fetching from it: https://launchpad.net/~ubuntu-acs/+livefs/ubuntu/trusty/stable/+build/53895 | 01:00 |
|---|---|---|
| lool | the build log has been on Ign https://private-ppa.launchpad.net trusty InRelease | 01:00 |
| lool | for a long time, then progressed by a couple of lines of the same type very slowly, and has been stuck for 10+mn or so | 01:01 |
| lool | alright, Connection timed out after 120001 milliseconds | 01:10 |
| lool | I guess it's a firewall issue | 01:10 |
| lool | I'll try passing this to the right admin group | 01:11 |
| cjwatson | lool: Sounds likely. I can't see the build to investigate further, since it's private. But that sounds plausible given that description. | 01:28 |
| sergio-br2 | hi | 01:48 |
| sergio-br2 | I'm having problem to build on arm64 | 01:48 |
| sergio-br2 | it's not finding dlopen for some reason | 01:49 |
| === maclin1 is now known as maclin | ||
| lool | cjwatson, wgrant: issue was tracked in a ticket and solved; firewall now lets private-ppa access through (over https) | 08:59 |
| === [HeOS] is now known as HeOS | ||
| clivejo | is there a problem with LP these days? Publishing seems to be taking a very long time compared to a few months ago? | 10:19 |
| cjwatson | clivejo: there's a known problem with bloat of a particular index due partly to the test rebuilds in progress. But can you give an example? | 10:49 |
| cjwatson | (Might or might not be the same thing, question is slightly too general to be able to tell) | 10:50 |
| clivejo | cjwatson: builds in kubuntu-ci seem to be taking a very long time to publish - https://launchpad.net/~kubuntu-ci/+archive/ubuntu/unstable | 10:51 |
| cjwatson | Can you quantify that? | 10:56 |
| cjwatson | Anywhere up to about 15 minutes is normal. | 10:56 |
| cjwatson | Maybe even 20, it depends where you happen to land in the cycle. | 10:56 |
| clivejo | we have a Jenkins instance monitoring it, some are taking up to 30mins | 10:57 |
| cjwatson | That can happen if there's a particularly long run. | 10:58 |
| clivejo | there are some builds that used to finish in 8mins now taking 20mins or more | 11:00 |
| cjwatson | The average cycle time at the moment seems to be around 15 minutes, so if your build happens to enter the queue just after the start of a publisher run, it could end up waiting for near the end of the next one if it's unlucky. | 11:00 |
| cjwatson | Ah, are you now talking about builds, not about publishing? | 11:00 |
| clivejo | just seems that there is a noticeable delay in this past few months | 11:00 |
| cjwatson | Some is just organic growth. | 11:01 |
| cjwatson | There is definitely a problem at the moment with collecting artifacts from builders at the end of a build, though, or at least was yesterday morning. | 11:01 |
| clivejo | well the timer is the entire process, but investigation seems to show that the publishing is the hold up | 11:01 |
| cjwatson | Publishing is always the slowest part. | 11:01 |
| cjwatson | (On average.) | 11:01 |
| clivejo | what is the reason for that? | 11:02 |
| cjwatson | But log analysis doesn't suggest anything particularly worse right now than e.g. a year ago. | 11:02 |
| clivejo | Im guessing there is a technical bottleneck here? | 11:02 |
| cjwatson | There are a lot of PPAs and the publisher has some linear scaling. | 11:02 |
| cjwatson | It's walking through all PPAs that need work each cycle, and there are a lot of those. | 11:03 |
| cjwatson | But it's also not structured to permit bulk operations very well. | 11:03 |
| cjwatson | Still, if you're seeing something in the last couple of days that's noticeably worse than a few months ago, which is how you opened this conversation, it's likely not actually the publisher. | 11:04 |
| cjwatson | (Also for the last day we've had logging turned up to 11 to try to get good data on buildd-manager database handling ...) | 11:09 |
| clivejo | cjwatson: can it depend on the PPA as well? | 11:10 |
| clivejo | for example my own PPA seems to publish a lot faster than the kubuntu-ci one | 11:11 |
| cjwatson | clivejo: kubuntu-ci is GIANT | 11:11 |
| cjwatson | clivejo: But it's more that giant PPAs like kubuntu-ci slow the whole cycle down | 11:11 |
| cjwatson | clivejo: It will depend on how you check | 11:12 |
| cjwatson | We can hardly complain about people using the service of course, but kubuntu-ci does use it extensively :-) | 11:12 |
| cjwatson | clivejo: If you're in a cycle where kubuntu-ci hasn't done anything but your own PPA has, then it's possible that that particular cycle will be quicker | 11:13 |
| clivejo | so the publisher is a batch job? | 11:15 |
| clivejo | does that mean the actual build machine is idle until the files are copied off? | 11:16 |
| cjwatson | The publisher is a batch job, yes; it tries to run every five minutes, though only actually runs if another instance isn't running. | 11:17 |
| cjwatson | Yes, but the builder doesn't wait on the publisher; retrieving files from the builder is a separate piece. | 11:18 |
| clivejo | interesting | 11:18 |
| HeOS | cjwatson, hello! We have a problem with our commercial subscription on Launchpad. Launchpad is returned the following ID: OOPS-6169ec49f0d826b02ef3cddb17dbd80b. Could you help us? | 11:51 |
| ubot5 | https://oops.canonical.com/?oopsid=OOPS-6169ec49f0d826b02ef3cddb17dbd80b | 11:51 |
| cjwatson | HeOS: can you try again now? We've started a proxy that wasn't running for some reason. | 12:02 |
| HeOS | cjwatson, thanks, I'm going to follow your recommendations. | 12:16 |
| === psivaa_ is now known as psivaa | ||
| === FourDollars_ is now known as FourDollars | ||
| === igitoor_ is now known as igitoor | ||
| === Guest86703 is now known as jelmer | ||
| sergio-br2 | hi | 23:34 |
| sergio-br2 | <sergio-br2> I'm having problem to build on arm64 | 23:34 |
| sergio-br2 | <sergio-br2> it's not finding dlopen for some reason | 23:34 |
| sergio-br2 | is this the right channel? | 23:34 |
| sergio-br2 | Checking function dlopen in -ldl ... no | 23:46 |
| sergio-br2 | isn't it from glibc lib? | 23:46 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!