[Buildbot-devel] buildbot fails after working for a while

[Daniel]

Greetings, all.

I have a couple of odd problems with Buildbot and wondered if anyone  
had seen these before and knew how to get past them.  I've done some  
Google searches and some other experimentation but haven't figured out  
how to get past these issues.

I recently upgraded from 0.7.5 to 0.7.7, and that's when both issues  
appeared.

Issue #1: I can't get the "Force Build" feature to work anymore.  In  
0.7.5, it worked fine.  In 0.7.7, "Force Build" does nothing no matter  
how I fill in the form.  I submit the form with my name, with  
"testing" as the reason, and "None" for both the branch and the  
revision.  I have tried other words for branch & revision too (such as  
"Main" and "Latest" or "*" for both fields), but nothing seems to  
work.  Any idea what I should do about that?  In 0.7.5, I was able to  
force a build, no problem.

Issue #2: When I upgraded to 0.7.7, I added a new build slave: OS X  
(Leopard).  What I see is that after working successfully for a while,  
the OS X slave will fail CVS checkout, with "You don't exist, go  
away!"  I have researched that error message, and it seems to be an  
SSH authentication issue (the comma in the error message gives that  
away).  The error doesn't make sense, though, because the exact same  
command works manually, and the slave itself works for a while and  
then just quits working - it builds fine for a while and on a  
subsequent build it starts generating that error message.  Restarting  
the slave doesn't always fix the problem, and restarting the master  
doesn't fix it either - sometimes the entire build system needs to be  
restarted.  Whether a restart fixes it seems to be random, but  
whenever the buildslave starts failing, it never works again and must  
be restarted, sometimes a number of times.

Any ideas on how I can resolve these?

Thanks.

Daniel