On Thu, May 01, 2014 at 11:20:00AM -0400, post...@nisny.com wrote: > My active queue keeps growing over the past 90 minutes (1410 messages at > this point). qmgr is not throwing any warnings, errors, panics, etc. > qshape reports domains affected as local, so appear to be incoming or > local emails. > > Reviewing the logs it appears that emails come into postfix and there is a > huge delay before sending to amavisd. Once cleared of amavisd they are > delivered promptly. > > master.cf: > smtp-amavis unix - - n - 10 lmtp > -o lmtp_data_done_timeout=1200 > -o lmtp_send_xforward_command=yes > -o disable_dns_lookups=yes > -o max_use=20
No need for disable_dns_lookups here. There is no reason to set max_use to 20. A process limit of 20 rather than 10 here might help. > amavisd IS configured for 10 processes and top shows them running but > mostly not being used so it doesn't appear that is a bottleneck. If amavisd is doing remote DNS lookups (RBLs, URIBLs, ...) then your 10 processes are likely stalled. Disable remote lookups or increase content filter concurrency. If you changed something recently, that change may have had performance consequences. Do measure the moving average "b" and "c+d" latency for messages hitting amavis, also measure the message output rate. Compare with numbers before the problem started. See recent related thread. -- Viktor.