When a RBL server your postfix is using goes off-line, postfix will
continue to query it, taking a lot longer, 10s of seconds, to accept each
incoming mail. Each SMTPD process lasts a lot longer and so if you hit
the default max of 50 SMTPD processes, incoming mail will be refused.
I had call yesterday from an IMGate user who said his mail volumes were way
down. The problem was osirusoft being unreachable. I removed osirusoft
from his smtpd restrictions about 19:00 and look at the results:
Per-Hour Traffic Summary
time received delivered deferred bounced rejected
--------------------------------------------------------------------
0000-0100 2117 393 9 4 1353
0100-0200 2216 431 13 2 1400
0200-0300 2220 263 5 2 1420
0300-0400 2303 235 3 1 1486
0400-0500 2261 237 3 0 1428
0500-0600 1867 226 3 2 1023
0600-0700 1813 321 1 2 857
0700-0800 1732 504 2 1 644
0800-0900 1788 459 7 3 692
0900-1000 1941 623 6 2 695
1000-1100 1835 749 14 5 494
1100-1200 2005 910 27 3 563
1200-1300 2162 1158 24 6 571
1300-1400 1810 655 27 5 605
1400-1500 1799 554 11 2 623
1500-1600 1997 667 17 6 674
1600-1700 1963 552 194 4 747
1700-1800 2091 616 19 1 830
1800-1900 1927 589 28 3 696
1900-2000 1891 793 22 1 578
2000-2100 7136 3271 4204 3
4036<osirus removed
2100-2200 7918 3841 383 13 5083
2200-2300 6491 2117 33 4 4886
2300-2400 6310 1324 18 5 5769
postfix opened up again to full speed, both for accepting and delivering
messages as well as rejecting messages.
( the huge number of deferrals 4200 had nothing to do with the osirus RBL,
but his IMGate was also screwed up and giving a lot of "connection refused"
to IMGate, so the I was repeatedly flush the queue as he tried to get IMail
to behave. )
For low-volume servers, a dead RBL is no big deal, but for higher volume
IMGates, it can really restrict.
I guess I need to work on some maillog scripts that monitor deferrals to
the mail box server (Imail not accepting mail) and monitor failures to
lookup RBL servers.
Len
