375156 - WINNT 5.1 qm-winxp01 dep unit test burning like it's bug 374409

Status: RESOLVED FIXED

(Release Engineering :: General, defect)

Description

[Phil Ringnalda (:philor)]

qm-winxp01 has fallen, and can't kill its process, so it needs someone to kick it while it's down.

robcee: how's your weekend so far?

Comment 1

[Dave Miller [:justdave]]

This doesn't look like bug 374409, there's no failure to remove a directory.

I'm having trouble getting into this box... none of the passwords I was given work.  robcee, any help?

Comment 2

[Phil Ringnalda (:philor)]

I didn't give the full explanation, since I was only planning on reminding robcee - bug 374409 wasn't really failure to remove, it was failure to kill a timed out process the first time, and then since that process was still hanging onto the directory, failure to remove the directory on subsequent builds, which seems to be the same thing the logs are showing now: one failure to kill, then failure to remove.

Comment 3

[Justin Fitzhugh]

This box is in the office and not even on the build documentation.  Preed/rhelmer - is this a tier 1 tinderbox (or even part of the farm)?  If not, should not be paged on it...if so, it should be added and moved to the colo.

Comment 4

[J. Paul Reed [:preed]]

(In reply to comment #3)
> This box is in the office and not even on the build documentation. 
> Preed/rhelmer - is this a tier 1 tinderbox (or even part of the farm)?  If not,
> should not be paged on it...if so, it should be added and moved to the colo.

It's not on the build documentation because it's not a build tinderbox.

It's a QA-managed box, and we need to help them get into the IT on-call infrastructure. Right now they're not, so yes... this shouldn't page you guys, and should be assigned to robcee.

Comment 5

[Justin Fitzhugh]

Putting in build/release for lack of a qa component, but assigning to robcee (of qa)

Comment 6

[Robert Sayre]

Yes, this has happened a couple of times.

Comment 7

[Dave Miller [:justdave]]

For the record I did get ahold of robcee and got access to the box and kicked it before finding out it wasn't our problem.

Comment 8

[Robert Sayre]

(In reply to comment #7)
> For the record I did get ahold of robcee and got access to the box and kicked
> it before finding out it wasn't our problem.
> 

Thanks Dave.

Comment 9

[Rob Campbell [:rc] (:robcee)]

I logged in and did a full clobber. The machine's still building an hour later. I'll check back in the morning to see if it's cleared up, and if not, try to track down the problem.

Comment 10

[Phil Ringnalda (:philor)]

61 minutes to greenness - it must have just been watching and waiting for you to give up on it. Thanks for the off-hours smack upside its head.

Comment 11

[Olli Pettay [:smaug][[e-mail]]]

And it is burning again...
The problem looks pretty similar

Comment 12

[Phil Ringnalda (:philor)]

Yeah, I did have a little "one good build might be jumping the gun" feeling when I closed this bug.

Comment 13

[Rob Campbell [:rc] (:robcee)]

the actual bug is a side-effect of *something* in JavaScript causing Mochitest to freeze which, on windows, buildbot isn't able to effectively terminate. We've been seeing this behavior for at least a couple of weeks now and until the root cause is fixed, or we patch buildbot, the problem will persist. Please be patient.