At 02:31 PM 2/11/99 +0100, Franky Van Liedekerke wrote:
>A small corrections: there are also a number of mails for local delivery (and
>forwarding).
>
>I witness the following behaviour:
>
>the first x mails go through without any problems (no problem with the
>concurrency: the number of remote processes is always less then 30).
>Then, at a sudden point, I see for all newly queued mails (the rest of the
>3000) in the logfiles:
>
>918738205.279362 delivery 14284: deferral:
>qmail-spawn_unable_to_create_pipe._(#4.3.0)/
>
>This goes on for all queued emails, and after that newly arrived emails go
>through again without problems...
>The same if I send an alarm to qmail-send.
>Am I hitting the concurrencylocal limit here? Do I need to reload or restart
>qmail for this parameter to take effect?
Nope. This is a classic problem. You have hit a resource limit imposed on
qmail by your startup script.
All it is saying is that qmail wants to fork another qmail-remote, but the
fork is failing due to exceeding a ulimit.
We need to know your OS, but in any event you need to increase the number of
child processess in the start-up script. Eg, with solaris, I go:
ulimit -n 256
prior to running qmail-start.
Obviously setting this limit is dependent on your OS.
Regards.
