[Buildbot-devel] problem "want to start build, but we don't have a remote"

[Stephen Davis]

The Windows slave I have running on XP randomly disconnects and  
reconnects.  My Windows 2003 server slave does not but they're  
running different Python/Twisted/buildbot versions.  I've never  
gotten around to figuring out which of the different bits is making  
the XP machine disconnect.

The master does notice though so this doesn't sound like your problem  
either.  I just thought I'd mention it in case anyone is seeing the  
same thing.  It appears to be specific to the combination of versions  
I have installed.  Windows slaves have always been, shall we say,  
temperamental.  :-)

stephen

On Jul 13, 2007, at 3:12 PM, Timothee Besset wrote:

> I agree there should be an upper limit on the progressive backoff if
> there isn't, but I don't think that's what the problem is. The slave
> still answers to ping checks through the web interface, so it's there
> and connected. For some reason, it just not starting builds. Only
> happens on Windows slave as well, the usual answer is just to push the
> big red button.
>
> TTimo
>
> Stephen Davis wrote:
>> On Jul 3, 2007, at 10:32 PM, Brian Warner wrote:
>>
>>
>>> Timothee Besset <ttimo at idsoftware.com> writes:
>>>
>>>
>>>> I'm getting a weird situation where one of the builders
>>>> accumulates the
>>>> build pending but never starts them. We always use the http
>>>> interface to
>>>> force the build to start, and now it just says 'n pending', 'n+1
>>>> pending'.
>>>>
>>> It kind of sounds like a reconfig took place and the new Builder
>>> failed to
>>> connect up with the existing slave. Maybe a race condition in the
>>> reconfig
>>> code?
>>>
>>
>> Could this be a case of the master being offline for long enough that
>> the build slave's "progressive backoff" has now reached a ridiculous
>> value?  The slave will reconnect *at some point* but I think it'd be
>> better if the progressive backoff had an upper limit (or, if it has
>> one now, that it be lowered).
>>
>> Or, just restart the build slave and see if it all hooks up
>> properly.  The master's log might tell you if the slave did try to
>> connect but failed.  The slave's log certainly should.
>>
>> stephen
>>