Am 12.03.21 um 08:27 schrieb Andrew C Aitchison via Exim-users:
Any ideas what's happening here ?
The increasing delays remind me of the delays for failed logins.
Is there some sort of authentication, or similar pam or SMB access,
to check the existence of each recipient ?
i made a test mail with 3000 recipients in CC.
Exim is increasing it's CPU Usage the more recipients are used until
100% CPU Core usage is reached:
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
1378404 root 20 0 66084 55300 10632 R 98,7 1,1 8:59.99
/usr/sbin/exim -Mc 1lKd2y-005m5p-3V
I have a static linear timeframe from one recipient to another appearing
in the log file, once the actual delivery from thunderbird ended.
strace says, it's due to my SQL select i use. it would go faster if the
SQL result would be cached, instead of repeating itself.
Conclusion:
a) the ( recipient ) listfunction in exim needs a redo, to improve
performance.
This is a very basic list optimization problem, you learn in data
structures at the university.
b) the setup in question has an additional problem, increasing the time
exponential .
because the core is reaching 100%, it's at it's max.. at that point,
things slow down even more.
c) time to make exim multi-threaded
best regards,
Marius
--
## List details at https://lists.exim.org/mailman/listinfo/exim-users
## Exim details at http://www.exim.org/
## Please use the Wiki with this list - http://wiki.exim.org/
