Description of problem: Currently if a build fails, I find the build directory is at least frequently -- maybe always? -- 404 afterwards, so you can't diagnose the failure. My latest example is for ppc64le under https://copr.fedorainfracloud.org/coprs/loveshack/livhpc/build/157659/
OK, it doesn't happen all the time. I repeated one case and did get a directory with the failed build results.
Hello, this is because failed builddirs get deleted pretty quickly (every day basically). We will work on that to keep failed build logs for a longer time. Thanks.
(In reply to clime from comment #2) > Hello, this is because failed builddirs get deleted pretty quickly (every > day basically). We will work on that to keep failed build logs for a longer > time. Thanks. OK, thanks. In case it's not meant to be as quick, I'd have said it happened on a timescale of order an hour, but that might be bad luck for building too close to an event. However, I wonder if something else is happening. There's also an example of a successful build that I've just noticed has disappeared somehow under https://copr.fedorainfracloud.org/coprs/loveshack/livhpc/package/libmsr/ (number 151920).
(In reply to clime from comment #2) > Hello, this is because failed builddirs get deleted pretty quickly (every > day basically). We will work on that to keep failed build logs for a longer > time. Thanks. Not true. Failed builds are preserved for 14 days.
(In reply to Miroslav Suchý from comment #4) > (In reply to clime from comment #2) > > Hello, this is because failed builddirs get deleted pretty quickly (every > > day basically). We will work on that to keep failed build logs for a longer > > time. Thanks. > > Not true. Failed builds are preserved for 14 days. How much do you bet? ;-) Failed builds _should_ be preserved for 14 days but the code that handles their deletion contains a bug which would result in them being preserved for just 14 seconds. The pruning script runs from cron.daily, however, so it is in order of hours until they get deleted.
No, there are currently lot of "spontaneous" build failures on Copr. Builds are often failing immediately on all roots. I guess before even any file has been written to the results dir. I think this bug should be renamed "frequent instant build failures".
(In reply to Jens Petersen from comment #6) > I think this bug should be renamed "frequent instant build failures". No. I was initially talking about cases where I'd looked at the result a little after a build failed and then went back for a look next day to fix it. There may be multiple bugs, of course
(In reply to Jens Petersen from comment #6) > No, there are currently lot of "spontaneous" build failures on Copr. > Builds are often failing immediately on all roots. > I guess before even any file has been written to the results dir. > > I think this bug should be renamed "frequent instant build failures". If the problem persists, let us know. Note that we have got quite a lot of problems with PPC64LE architecture (insufficient building resources). Also building from GIT or other SCMs didn't work correctly till recently. The situation might be better now.
Fixed by a5c93116e2 (https://github.com/fedora-copr/copr/). It might take a while till it gets into production. Currently, prunning is off.
*** This bug has been marked as a duplicate of bug 1304012 ***