On Thu, Sep 25, 2014 at 09:59:44PM -0700, Will Yardley wrote:
> I will try running strace on an smtpd process and see if that helps
> identify anything.
Notice the nanosleep() call here, and the jump from 23:17:21.642640 to
23:17:22.642808 -- exactly a second.
23:17:21.608881 read(9, "RCPT TO:<xx...@example.com>\r\n", 4096) = 31 <0.000009>
[...]
23:17:21.642026 read(13,
"flags\0000\0transport\0local\0nexthop\0XXXX.example.com\0recipient\0xx...@example.com\0flags\000256\0\0",
4096) = 92 <0.000012>
[...]
23:17:21.642517 flock(21, LOCK_UN) = 0 <0.000007>
23:17:21.642546 rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0 <0.000008>
23:17:21.642582 rt_sigaction(SIGCHLD, NULL, {SIG_DFL, [], 0}, 8) = 0 <0.000008>
23:17:21.642615 rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0 <0.000008>
23:17:21.642640 nanosleep({1, 0}, 0x7fffbf08c600) = 0 <1.000107>
23:17:22.642808 flock(23, LOCK_SH) = 0 <0.000010>
23:17:22.643090 flock(23, LOCK_UN) = 0 <0.000008>
$fork_delay wouldn't affect this here, would it (it's set to 1s, but
can't be set lower anyway)? I re-tried setting in_flow_delay and did a
postfix reload, but that didn't change anything.
w
