On Mon, Mar 31, 2014 at 10:55:04PM +0530, KK Patnaik wrote:
> 1) Bounces are for outbound sent to lot of email addresses. And this is not
> spamming.
Many bulk email senders believe the spammers are all the other bulk
senders, but not they.
> These are only outbound servers.
> 2) qshape -s
Why "-s", this gives no information about the destination of slow
email. You should probably also scan the deferred queue, thus:
# qshape incoming active deferred
If the output device is a terminal, qshape will give you progressive
output for every 1000 messages found.
> The qshape information is around an hour back information. Now for last
> 40mins it's still at the command prompt and I am not seeing any output as of
> now.
Sounds like you got slammed with a bunch of new mail and your disk
sub-system in not fast enough for qshape to read the entire queue
in a timely manner, especially with mail processing competing.
> 3) Can you please suggest any tool which can give me ASCII histogram for the
> emails for the "c+d" delays.
You write a Perl or Python script to parse this from delays=a/b/c/d
in your own logs.
> bounce_queue_lifetime = 0
> maximal_queue_lifetime = 0
Bad idea.
> non_smtpd_milters = unix:/var/run/dkim-milter/dkim-milter.sock,
> unix:/var/run/dk-milter/dk-milter.sock
> smtpd_milters = unix:/var/run/dkim-milter/dkim-milter.sock,
> unix:/var/run/dk-milter/dk-milter.sock
Is this keeping up with the mail stream? Perhaps this is a bottleneck.
You must read your logs and determine what's going on. Have you tried
the "collate" script I've sent?
> Can you please suggest how to address the issue now?
The data is in your logs, understanding them well enough to identify
the problem source is your responsibility.
You need to determine whether your CPU, network, disk or output
concurrency are maxed out, or whether remote destinations are
throttling your deliveries, ...
Perhaps you syslogd is misconfigured and is logging synchronously.
See LINUX_README.html. That would cause the disk to be swamped.
--
Viktor.
