Comment 2 for bug 503440

Michael Nelson (michael.nelson) wrote :

I've just checked the IRC logs, and it looks likely that the above last dispatch on 2009-12-21 18:20:11 was the cause of a complete build farm failure (bug 499421), as there was some SQL run at that point:

https://irclogs.canonical.com/2009/12/21/%23launchpad-code.html#t19:35 (sorry for the non-open link)

which did:

{{{
update job set date_started = null
FROM
    build, buildqueue bq, buildpackagejob bpj, archive
WHERE
    bq.job = bpj.job AND
    bpj.build = build.id AND
    build.archive = archive.id AND
    bq.builder is null AND
    job.id = bpj.job AND
    job.date_started is not null;
}}}

as setting the date_started back to null allowed the build farm to continue in a healthy state. But it does not leave the build in a healthy state. I assume (it would have been good to have had the build id) it is this build that was the culprit, and it had it's date_started set back to null, but the buildstate was still BUILDING - hence this oops.