Hi,
We detected high load on a server, and looks like two qpsmtpd-forkserver
processes were munging up the CPU, and looks like they had been there
for a long time :S
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
7936 user001 25 0 21216 16m 3236 R 67 0.8 26151:31
/usr/bin/qpsmtpd-forkserver [195.232.246.84 : mo1.vodafone.com :
15:57:41 2008-03-10]
11656 user001 25 0 21216 16m 3236 R 38 0.8 26106:25
/usr/bin/qpsmtpd-forkserver [195.232.246.81 : mef1.vodafone.com :
16:55:59 2008-03-10]
I attached an strace to each process and got that they where
continuously (many times per second) doing this operation:
on PID 7936:
--- SIGPIPE (Broken pipe) @ 0 (0) ---
write(8, "\27\3\1\0000\367\377\236h]_\237H\227\230q9\315\220\255"...,
53) = -1 EPIPE (Broken pipe)
on PID 11656:
--- SIGPIPE (Broken pipe) @ 0 (0) ---
write(8, "\27\3\1\0000\25\252\234)\323]\344\0~t\200\212\334\273\215"...,
53) = -1 EPIPE (Broken pipe)
Had to kill them off to stabilize the situation. Later I realized that I
should have done a lsof on the to see what the FD 8 was supposed to be
:S. Looking at other qp processes on the same machine, FD 8 on qp
processes seems to change roles:
qpsmtpd-f 12964 user001 8u IPv4 440886485 UDP *:50420
qpsmtpd-f 12996 user001 8u IPv4 440885640 TCP
server_ip:smtp->remote_ip:12026 (ESTABLISHED)
Running qp 0.40 with forkserver on Debian Etch. Relevant plugins are tls
and qmail-queue. Any ideas on how to find the guilty plugin? Just seems
like it's ignoring SIGPIPEs wh en writing. And what could cause it?
Funny thing is that the two connections where from the same provider O_O.
Any ideas?
Jose Luis Martinez
[EMAIL PROTECTED]
