--On Thursday, February 19, 2009 1:54 PM -0500 Wietse Venema
<wie...@porcupine.org> wrote:
Quanah Gibson-Mount:
I've been running into a really odd (bizarre) problem with Postfix that
only seems to happen on Mac OSX 10.5 (leopard). I'm really at a loss to
explain why things break the way they do, but it definitely happens. I
thought maybe some folks on the list might have some insight.
Here is the scenario: Client installs Zimbra on their server. Zimbra
has its own Postfix build that it installs, with its own spool
directory (/opt/zimbra/data/postfix/spool) and its own configuration
files (/opt/zimbra/postfix/conf) etc. All the configuration files we
ship only refer to our locations. When ZCS is up and running,
everything on the system uses our postfix.
Now, what we've seen is that if ZCS is not up and running and the ZCS
user (zimbra) is not in existence, then the OSX Leopard box will fall
back to its local postfix, which of course puts files in
/var/spool/postfix. This in and of itself is not really a problem.
The problem is, that once ZCS is back up and running, if there are
files in /var/spool/postfix, eventually proxymap will stop being able
to talk to the LDAP server:
Feb 19 03:22:32 mx1 postfix/proxymap[53103]: error: dict_ldap_connect:
Unable to set STARTTLS: -1: Can't contact LDAP server
Removing the files from /var/spool/postfix makes this stop happening. I
can't for the world think of why files existing in /var/spool/postfix
would have any effect on our postfix which has no knowledge of that
location, but cleaning out /var/spool/postfix always resolves the issue.
Speculation: as long as the MacOSX Postfix queue is not empty, some
bits and pieces of MacOSX Postfix will keep running and occupy some
resource that your Postfix would otherwise have provided.
We disable the MacOSX postfix when we install ZCS. So the postfix
processes at the system level themselves are not running (and thus why the
mails just end up in /var/spool). It's simply that cron ends up using the
system /usr/bin/sendmail, and because ZCS isn't available, it spools it in
the wrong location. I've checked ps output, and there's definitely nothing
but our postfix running.
Anyone have an insight into why? Postfix version is 2.4.7.
This is really a platform-specific question, that can be answered
only by people who have access to the affected OS.
Ok, but it's also something directly related to postfix. Restarting the
postfix process will let proxymap run again for a while, until the STARTTLS
failures show up again. Cleaning out the /var/spool location is the only
thing that permanently resolves it.
--Quanah
--
Quanah Gibson-Mount
Principal Software Engineer
Zimbra, Inc
--------------------
Zimbra :: the leader in open source messaging and collaboration
