[Buildbot-devel] twisted.internet.tcp.Connector Uninitialized problems

[Bob Proulx]

Jean-Paul Calderone wrote:
> Bob Proulx wrote:
> >  2007/06/22 17:17 MDT [-] Starting factory <buildbot.slave.bot.BotFactory instance at 0x405b87ac>
> >  2007/06/22 17:17 MDT [Uninitialized] <twisted.internet.tcp.Connector instance at 0x405bf0ac> will retry in 71 seconds
> >  2007/06/22 17:17 MDT [Uninitialized] Stopping factory <buildbot.slave.bot.BotFactory instance at 0x405b87ac>
>
> The underlying TCP connection attempt is failing for some reason
> (not a reason that is getting logged, unfortunately).  I'd try
> debugging things at that level first - make sure you can reach the
> master address/port with another program (eg, telnet) and/or do a
> traffic dump (ethereal, tcpdump) on the machine when the slave makes
> its connection attempt.

Yes!  Thank you very much for that hint.  I was in disbelief about a
particular thing and your suggestion caused me to evaluate my thinking
again and was able to find the problem.

> You might find the connection is being blocked or failing for some
> simple reason.

Previous to posting my question I had double checked firewall rules.
I had double checked connection capability.  I was able to connect
using 'telnet', 'connect', etc. and could connect okay to the
buildmaster host.  When I read your suggestion I was of course
thinking that couldn't be it.  But with your analysis I knew that must
be the problem and it had to be in there somewhere and looked
everything over again with that in mind.  The problem when I found it
was user error, as probably expected by most of you. :-)

I had copied the buildbot.tac configuration from a buildbot slave
running on the buildmaster host itself.  This was the roundabout root
cause of the problem.  If I had picked another buildslave I would not
have tripped into this problem.  I then changed all occurrences of the
previous hostname to the new hostname in the configuration file.  This
is where my problems began.  The 'host' config there had the name of
the buildmaster and was also the name of the previous host.  It was
changed by my edit to the new hostname.

The new buildbot slave was trying to connect to a buildmaster on
itself.  Since none was running there it was getting connection
refused (it would be nice if that were logged) and failing.  When I
reviewed the file I saw 'host' there and the local hostname and it did
not stand out as obviously wrong.  (Naming it 'masterhost' instead of
'host' would help in the future.)  Argh.  I now feel silly to have
made such an error.

Thank you very much for that hint!

Bob