Обсуждение: Fwd: PGBuildfarm member colugos Branch HEAD Status changed from OK to StartDb-C:3 failure

Robert Creager <rsc@logicalchaos.org> writes:
> If anyone is interested, I think this failure was accompanied by the following:
> [ apparent PANIC in UpdateControlFile ]

Hmm, do you have the panic message in the postmaster log?  So far as I
can tell, the postmaster log isn't captured anywhere in the buildfarm
report of this event.  (Which seems like a buildfarm bug.)

Given that polecat has already run a couple of later buildfarm
iterations, I'm not too hopeful that you do have that log file.
Was there any special environment here, like running out of disk space?
        regards, tom lane

Robert Creager <robert@logicalchaos.org> writes:
> On May 20, 2010, at 11:54 AM, Tom Lane wrote:
>> Was there any special environment here, like running out of disk space?

> Not that I'm aware of.  I did empty trash sometime yesterday after noticing I was around 1Gb of free disk.  Not sure
ifthat correlates or not.  Maybe on failure buildfarm should could disk usage of the disk it's on and add that to the
report?

I'm just guessing, but it seems like a system-level condition like being
out of disk space would explain concurrent similar failures for both of
your animals.

A quick look at UpdateControlFile shows that it definitely will PANIC
if it gets any sort of failure while trying to write pg_control.
It would be interesting to know what sort of failure it got, but
I'm not going to (ahem) panic about it.  It's annoying though that
the buildfarm script didn't capture the relevant log file in this
particular failure case.  Andrew, can we get that fixed?
        regards, tom lane

On May 20, 2010, at 11:54 AM, Tom Lane wrote:

> Robert Creager <rsc@logicalchaos.org> writes:
>> If anyone is interested, I think this failure was accompanied by the following:
>> [ apparent PANIC in UpdateControlFile ]
>
> Hmm, do you have the panic message in the postmaster log?  So far as I
> can tell, the postmaster log isn't captured anywhere in the buildfarm
> report of this event.  (Which seems like a buildfarm bug.)

'Course not.

>
> Given that polecat has already run a couple of later buildfarm
> iterations, I'm not too hopeful that you do have that log file.
> Was there any special environment here, like running out of disk space?

Not that I'm aware of.  I did empty trash sometime yesterday after noticing I was around 1Gb of free disk.  Not sure if
thatcorrelates or not.  Maybe on failure buildfarm should could disk usage of the disk it's on and add that to the
report?

I've updated my OSX clients to keep error builds.

Later,
Rob

On Thu, May 20, 2010 2:12 pm, Tom Lane wrote:
> It's annoying though that
> the buildfarm script didn't capture the relevant log file in this
> particular failure case.  Andrew, can we get that fixed?
>


It was captured, but apparently had no new content - not surprising if it
ran out of space.

cheers