[Buildbot-devel] Questions on basic BuildBot configuration

[Mary Fernandez]

Greetings,

I have two very basic questions about BuildBot
configuration problems that hopefully someone 
can answer easily.

Thanks for any help in advance!
-Mary

General config info:
buildbot 0.7.5
Twisted 2.0.1   
CVSToys 1.0.10
(CVSToys did not compile against Twisted 2.5, so I backed
up to 2.0, which was in the same era as the CVSToys release)

Problem #1: SMTP error when using mail.mailNotifier
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Configuration: FreshCVS, a buildbot master, and
a buildbot slave are all running on the same machine.

What happens: On a commit, FreshCVS sends mail to 
me (successfully) and also triggers a PBService event, 
which in turn, triggers the build master, which contacts
the slave.  The slave runs the job, returns the result to the build
master, which attempts to send a mail notification. 

Results in following error in twistd.log:

 [Uninitialized] SMTP Client retrying server. Retry: 1
 [Uninitialized] Unhandled error in Deferred:
 [Uninitialized] Traceback (most recent call last):
	Failure: twisted.internet.error.ConnectionRefusedError: 
        Connection was refused by other side: 111: Connection refused.

I was surprised, because FreshCVS's email succeeded
so I expected BuildBot's to as well.

The difference is that FreshCVS's mail notifier 
invokes /usr/bin/sendmail explicitly as a shell command,
whereas BuildBot uses the twisted.mail.smtp object.

I'm guessing that sendmail on this machine is configured
in some way that ends up refusing connections from Twisted.
Has anyone ever seen this before?  Any suggestions?

Problem #2: Slave inside firewall cannot connect to master
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Configuration: FreshCVS and buildbot master are on machine
outside firewall.  buildbot slave is on machine inside
firewall (behind NAT)

What happens: I can SSH from slave machine to master
machine, check out source, and build successfully on the
slave.   But when I try to start the slave daemon, it never
connects to the master:

 [Uninitialized] <twisted.internet.tcp.Connector instance at 0xb7b3e18c>
will retry in 2 seconds

Running buildbot debugclient gives a bit more info:  

  [Failure instance: Traceback (failure with no frames): 
  twisted.internet.error.ConnectError: An error occurred while   
  connecting: 113: No route to host.

I'm guessing that this is the error when the daemon starts up.
There's obviously a route to the host, because SSH works.
Could this be due to incorrect address translation from the 
master back to the slave?  

Any ideas are greatly appreciated.
-- 
Mary Fernandez <mff at research.att.com>
AT&T Labs - Research