Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] Hudson shutdown wait from hell

On 02/10/2012 08:49 AM, David M Williams wrote:
>> ...  If you need a
>> job pronto, Fastlane is there for you.  That's what we set it up for.
>> It's currently idle.
> But "fastlane" often has the same "remote operation" failure as slave
> machines often do,

Right.  The master executes everything locally.  However, tying to the
master does not scale.

In the past we've tried creating one big master instance with tons of
executors and that didn't work out well either, so we have to use
slaves.  So no, you're not missing anything: at the moment you must pick
one solution from two sub-optimal choices:
-> tie to master, your builds succeed consistently, but you must wait in
line
-> don't tie to master, your build may fail because of remote operation
exceptions, but your build will complete faster

The third option, the one where your build always succeeds without ever
waiting, simply doesn't exist at this time.



On 02/10/2012 09:14 AM, Campo, Christian wrote:
> I am personally very much in favor of having a new mailing list for hudson
> build infrastructure messages. Many issues like slaves hanging, hudson is
> restarted are short-lived issues that are not always worth a bugzilla
The typical Hudson errors should be pretty well known by everyone at
this time, so when a slave has failed, why not just email webmaster@ and
keep it off the list?

Denis


Back to the top