On Fri, Dec 07, 2007 at 11:54:26AM -0500, David Mackintosh wrote:
I've recently replaced a RedHat EL 3.x system with a CentOS 5 system (fully yum'd as of Tuesday). This was a full-pave install, although we did copy the sendmail.mc from the original system.
Now I get a lot of this in my logs:
Dec 7 11:47:38 mail sendmail[20117]: lB7Gl6w0020116: timeout waiting for input from local during Draining Input
The only thing even remotely credible is a Sendmail Known Bugs page which suggests it is a chatty local delivery agent, but since this is happening with outbound messages I don't think I believe this.
Does anyone have any ideas what Sendmail is trying to tell me here?
Further to this, I have discovered that what Sendmail is trying to tell me is "procmail and dovecot are not playing nicely together".
What is happening is that I have a large amount of mail coming and going, and when procmail tries to deliver to a mailbox (in mbox format, in /var/spool/mail/$user, currently 12MB in size but can grow MUCH larger) that dovecot is actively using, procmail blocks -- and so do all the other procmail processes waiting to deliver to this mailbox. Sendmail gets annoyed waiting for these blocked procmail processes, and so fills my logs with the error message above.
If I do a "service dovecot stop", then the queued procmail processes all drain their messages into the affected mailbox, and the problem goes away for a little while after dovecot is restarted.
Naturally, the hack of stopping dovecot every so often isn't really a solution.
So the question is: how do I get dovecot and procmail to play nice? Is the solution to change to a maildir type inbox spool?