Strange errors from mailgate

Lately, I’ve been seeing these in my mail queue:

88E86B860 53247 Wed Nov 30 14:05:57 MAILER-DAEMON
(temporary failure. Command output: An Error Occurred
================= 500 Internal Server Error)

Basically, the mailgate is getting the "500 Internal Server Error"
from the REST interface to the web server, and reporting a tempfail
to the mail server (Postfix).

However, the big problem here is that the web server succeeded in
creating that ticket. So now when the mail server retries to give
the message to rt-mailgate, it will create yet another ticket.

There are no errors in the web server log for the corresponding time
from the mail server logs of when this error happened.

A similar thing happened a few days ago and cleared itself up after
stopping/restarting the web server. Today this did not help.

RT 3.4.4 running under FastCGI in Apache 2 on FreeBSD 6.0 with Perl
5.8.7 all installed from FreeBSD ports at the start of November.

Any clues would be appreciated.

Lately, I’ve been seeing these in my mail queue:

88E86B860 53247 Wed Nov 30 14:05:57 MAILER-DAEMON
(temporary failure. Command output: An Error Occurred
================= 500 Internal Server Error)

Basically, the mailgate is getting the "500 Internal Server Error"
from the REST interface to the web server, and reporting a tempfail
to the mail server (Postfix).

However, the big problem here is that the web server succeeded in
creating that ticket. So now when the mail server retries to give
the message to rt-mailgate, it will create yet another ticket.

There are no errors in the web server log for the corresponding time
from the mail server logs of when this error happened.

A similar thing happened a few days ago and cleared itself up after
stopping/restarting the web server. Today this did not help.

I’m betting this is a timeout. try upping the --timeout on your
rt-mailgate commandline?

I’m betting this is a timeout. try upping the --timeout on your
rt-mailgate commandline?

Oooohhhh. Looks like FastCGI has a 120 second idle timeout vs. RT’s
180 seconds… maybe that’s the problem. I’m making them the same to
see what happens.

Thanks for the pointer.