On 7 Dec 2006, at 08:13, Adrian von Bidder wrote:
Yuri, I'd appreciate your help here. I haven't heard back anything
from the
IO::Multiplex people, and I don't run exim, so I can't really test it.
With postfix, I've never seen this warning.
As mentioned, I don't want to add a logcheck rule because clearly the
warning should be handled, not hidden.
This happens when exim disconnects before the answer is written to
the socket, because I have a relatively short timeout (5 seconds). In
fact, this warning happens often just after postgrey has done some
kind of maintenance (performing maintenance before writing the
result????), but I'm having an hard time reproducing it "on demand".
Silencing the warning would be as simple as doing a kill_output when
Multiplex shuts down the connection, however I see no callback to
perform this _before_ the close (and the warning), and not after. I
guess that if postgrey was killed during the write it would output
the same warning.
I don't know, I don't have experience with IO::Multiplex nor enough
time to debug the problem.
I see an increasing amount of warnings, which match the increased
workload of the system.
postgrey taking as much as 5 seconds to write a result on a mildly
loaded system however isn't acceptable.
And performing maintenance _during_ the conversation is almost making
me reconsider the switch to postgrey.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]