no Job ID was provided in the time required (Was: Please rebuild your packages in Fedora Extras development)

[Ralf Corsepius]

On Wed, 2006-03-01 at 09:44 -0500, Dan Williams wrote:
> On Wed, 2006-03-01 at 13:32 +0100, Ralf Corsepius wrote:
> > On Wed, 2006-03-01 at 05:50 -0500, Josh Boyer wrote:
> > > On Wed, 2006-03-01 at 06:38 +0100, Ralf Corsepius wrote:
> > > > 
> > > > Either these issues can be nailed down in a reasonable manner in a
> > > > reasonable time frame, or other consequences will have to be taken ...
> > > 
> > > Like what?
> > 
> > Left to your imagination - There are many possibilities :-)
> > 
> > When, you, like me, have to experience 7 out of 30 buildjob die, because
> > the buildsystem chokes in different ways, and then have to experience
> > its maintainers telling you "everything is OK", despite you have the log
> > mails in front of you telling you else, as it happened to me last night,
> > you also will raise this question.
> 
> I still don't know what exactly "went wrong" with your builds...  

Remark: Last night I launched "2 batches of rebuilds", the first one
consisting of 9 jobs, the second one consisting of 21 jobs.

>From the first batch, I received 2 "no job ids", and observed the jobs
to be finishing with a delay of 5-6 hours. ATM, I presume you had
requeued them.

>From the 2nd batch 5 failed, with the symptoms matching your description
from your other mail ("full ppc disk/corrupted repo").
I requeued them, which hat let 4 of them succeed, while the 5th one
tripped a real package issue.

> >From what I can tell, there are a couple possiblities:
> 
> 1) You got "no Job ID provided"; but in this case the buildsystem still
> took your builds, and processed them.  I don't see any indication
> otherwise, please correct me if the builds did not get picked up by the
> buildsys at all
I don't know and can't tell. As mentioned above, the build jobs exposing
this symptoms completed with several hours delay.

> 2) Repodata was inconsistent last night for a while because mirrors were
> syncing.  That's not really the buildsystem's fault, though the buildsys
> could be more tolerant of that error.  When we install the depsolving
> upgrade over the next week or two, this problem should disappear as jobs
> that this this problem won't even get pushed to builders until its fixed
> 
> 3) ppc1 was out of disk space; I cleaned it up last night and requeued
> all the failed jobs I could find that had failed due to this error.

Cf. above, these 2 reasons definitely had been involved for the latter
5.

> I see two jobs of yours that have actually failed:
> 
> 5135: Coin2 (Coin2-2_4_4-3_2_fc4)  rc040203 at freenet.de   failed/failed
>         ppc1.fedora.redhat.com(ppc): 0ad02077f4fdbfab353a083c195e65c645705d10 done/prepping
>         hammer2.fedora.redhat.com(i386): 8e4d2c310ce60eda43c23e51b61784947370ffc1 done/failed
>         hammer3.fedora.redhat.com(x86_64): 48f4dbb01268431f3f91f1f01d00e41353882860 done/failed
> 
> ---> failed due to the dap-server issues last week; ultimately not buildsys-related
Correct, this is an older one, which was victim of the dap-server issues.

> 
> 5571: perl-DBIx-SearchBuilder (perl-DBIx-SearchBuilder-1_38-2_fc5)  rc040203 at freenet.de   failed/failed
>         hammer3.fedora.redhat.com(noarch): 6785ca3835d57e7d3753e70823bfb707574f986b done/failed
> 
> ---> Failed due to build errors, not buildsystem related
Also correct. This is the 5th one mentioned above.

Ralf