Mahout builds

[Antoine Toulme <an...@lunar-ocean.com>]

Two mahout builds were taking over vesta.

I killed both, and checked that they both have timeouts.
They both have one, but they don't have the checkbox "Fail the build" under
it checked.

It looks like to my eyes this checkbox is necessary to kill the build and
stop recording stuff.

Is it a fair assumption ?

Thanks,

Antoine

Re: Mahout builds

[Niklas Gustavsson <ni...@protocol7.com>]

On Tue, Aug 24, 2010 at 10:36 PM, Antoine Toulme
<an...@lunar-ocean.com> wrote:
> I discussed with the kind folks in #hudson. In both cases, the build is
> killed.
> If the button is checked, the build will show as failed (red ball),
> otherwise it will be blank (white/grey ball).

Confirmed from inspecting the code for the plugin.

We currently set this to be disabled in the timeout script, I'll
change that so that we don't touch the setting.

/niklas

Re: Mahout builds

[Antoine Toulme <an...@lunar-ocean.com>]

+1

On Tue, Aug 24, 2010 at 15:13, sebb <se...@gmail.com> wrote:

> On 24 August 2010 21:36, Antoine Toulme <an...@lunar-ocean.com> wrote:
> > I discussed with the kind folks in #hudson. In both cases, the build is
> > killed.
> > If the button is checked, the build will show as failed (red ball),
> > otherwise it will be blank (white/grey ball).
>
> In that case the "fail" help could do with updating. It says:
>
> "If checked, the build will be failed rather than aborted."
>
> That could easily mean that the build will not be aborted at all.
>
> I'll see about raising a bug report.
>
> > So it's up to projects to consider builds that take too long as failures.
> > For Buildr I consider a build that goes over the timeout to be a failed
> > build for example.
> >
> > On Tue, Aug 24, 2010 at 12:18, Niklas Gustavsson <niklas@protocol7.com
> >wrote:
> >
> >> On Tue, Aug 24, 2010 at 5:47 PM, Antoine Toulme <
> antoine@lunar-ocean.com>
> >> wrote:
> >> > Two mahout builds were taking over vesta.
> >> >
> >> > I killed both, and checked that they both have timeouts.
> >> > They both have one, but they don't have the checkbox "Fail the build"
> >> under
> >> > it checked.
> >> >
> >> > It looks like to my eyes this checkbox is necessary to kill the build
> and
> >> > stop recording stuff.
> >>
> >> While I would have agreed with your interpretation. However, as
> >> mentioned by sebb, the checkbox needs to be unchecked. Else, Hudson
> >> will only fail the build, not attempt to kill it.
> >>
> >> Sometimes, even with the build timeout, Hudson fails to kill stuck
> >> builds. One reason seems to be when the socket to a slave is broken.
> >>
> >> /niklas
> >>
> >
>

Re: Mahout builds

[sebb <se...@gmail.com>]

On 24 August 2010 21:36, Antoine Toulme <an...@lunar-ocean.com> wrote:
> I discussed with the kind folks in #hudson. In both cases, the build is
> killed.
> If the button is checked, the build will show as failed (red ball),
> otherwise it will be blank (white/grey ball).

In that case the "fail" help could do with updating. It says:

"If checked, the build will be failed rather than aborted."

That could easily mean that the build will not be aborted at all.

I'll see about raising a bug report.

> So it's up to projects to consider builds that take too long as failures.
> For Buildr I consider a build that goes over the timeout to be a failed
> build for example.
>
> On Tue, Aug 24, 2010 at 12:18, Niklas Gustavsson <ni...@protocol7.com>wrote:
>
>> On Tue, Aug 24, 2010 at 5:47 PM, Antoine Toulme <an...@lunar-ocean.com>
>> wrote:
>> > Two mahout builds were taking over vesta.
>> >
>> > I killed both, and checked that they both have timeouts.
>> > They both have one, but they don't have the checkbox "Fail the build"
>> under
>> > it checked.
>> >
>> > It looks like to my eyes this checkbox is necessary to kill the build and
>> > stop recording stuff.
>>
>> While I would have agreed with your interpretation. However, as
>> mentioned by sebb, the checkbox needs to be unchecked. Else, Hudson
>> will only fail the build, not attempt to kill it.
>>
>> Sometimes, even with the build timeout, Hudson fails to kill stuck
>> builds. One reason seems to be when the socket to a slave is broken.
>>
>> /niklas
>>
>

Re: Mahout builds

[Antoine Toulme <an...@lunar-ocean.com>]

I discussed with the kind folks in #hudson. In both cases, the build is
killed.
If the button is checked, the build will show as failed (red ball),
otherwise it will be blank (white/grey ball).

So it's up to projects to consider builds that take too long as failures.
For Buildr I consider a build that goes over the timeout to be a failed
build for example.

On Tue, Aug 24, 2010 at 12:18, Niklas Gustavsson <ni...@protocol7.com>wrote:

> On Tue, Aug 24, 2010 at 5:47 PM, Antoine Toulme <an...@lunar-ocean.com>
> wrote:
> > Two mahout builds were taking over vesta.
> >
> > I killed both, and checked that they both have timeouts.
> > They both have one, but they don't have the checkbox "Fail the build"
> under
> > it checked.
> >
> > It looks like to my eyes this checkbox is necessary to kill the build and
> > stop recording stuff.
>
> While I would have agreed with your interpretation. However, as
> mentioned by sebb, the checkbox needs to be unchecked. Else, Hudson
> will only fail the build, not attempt to kill it.
>
> Sometimes, even with the build timeout, Hudson fails to kill stuck
> builds. One reason seems to be when the socket to a slave is broken.
>
> /niklas
>

Re: Mahout builds

[Niklas Gustavsson <ni...@protocol7.com>]

On Tue, Aug 24, 2010 at 5:47 PM, Antoine Toulme <an...@lunar-ocean.com> wrote:
> Two mahout builds were taking over vesta.
>
> I killed both, and checked that they both have timeouts.
> They both have one, but they don't have the checkbox "Fail the build" under
> it checked.
>
> It looks like to my eyes this checkbox is necessary to kill the build and
> stop recording stuff.

While I would have agreed with your interpretation. However, as
mentioned by sebb, the checkbox needs to be unchecked. Else, Hudson
will only fail the build, not attempt to kill it.

Sometimes, even with the build timeout, Hudson fails to kill stuck
builds. One reason seems to be when the socket to a slave is broken.

/niklas

Re: Mahout builds

[sebb <se...@gmail.com>]

On 24 August 2010 19:13, Drew Farris <dr...@apache.org> wrote:
> Antoine,
>
> Thanks for pointing this out. I checked the 'fail the build'
> checkboxes and updated the configs for a couple of the jobs (before
> hudson started having issues). I'm assuming it's the build timeout
> plugin that takes care of killing the jobs. According to the docs[1]
> builds might still get need to be killed depending on how they're
> getting stuck. What was the name of the job(s) that were stuck?

I thought the "fail" option meant that the build would be marked as
failed if it took too long.
Whereas if the "fail" option is not selected, the build will be killed
if it takes too long.

However I cannot find any documentation on this.

> Drew
>
> [1] http://wiki.hudson-ci.org/display/HUDSON/Build-timeout+Plugin
>
> On Tue, Aug 24, 2010 at 11:47 AM, Antoine Toulme
> <an...@lunar-ocean.com> wrote:
>> Two mahout builds were taking over vesta.
>>
>> I killed both, and checked that they both have timeouts.
>> They both have one, but they don't have the checkbox "Fail the build" under
>> it checked.
>>
>> It looks like to my eyes this checkbox is necessary to kill the build and
>> stop recording stuff.
>>
>> Is it a fair assumption ?
>>
>> Thanks,
>>
>> Antoine
>>
>

Re: Mahout builds

[Drew Farris <dr...@apache.org>]

Antoine,

Thanks for pointing this out. I checked the 'fail the build'
checkboxes and updated the configs for a couple of the jobs (before
hudson started having issues). I'm assuming it's the build timeout
plugin that takes care of killing the jobs. According to the docs[1]
builds might still get need to be killed depending on how they're
getting stuck. What was the name of the job(s) that were stuck?

Drew

[1] http://wiki.hudson-ci.org/display/HUDSON/Build-timeout+Plugin

On Tue, Aug 24, 2010 at 11:47 AM, Antoine Toulme
<an...@lunar-ocean.com> wrote:
> Two mahout builds were taking over vesta.
>
> I killed both, and checked that they both have timeouts.
> They both have one, but they don't have the checkbox "Fail the build" under
> it checked.
>
> It looks like to my eyes this checkbox is necessary to kill the build and
> stop recording stuff.
>
> Is it a fair assumption ?
>
> Thanks,
>
> Antoine
>