> mmm, 10+ minutes elapsed - and 57448 has neither timed out nor
> noticed that 57451 died. Or sent status, but then it should
> have said *somthing* more. I wonder if it crashed (as seen in
> one previous event) and 57448 didn't see it?
I have the same exact debug output from exim d+all -M message_id on my FreeBSD 13 servers.
Those stuck processes will hang around for days if you don’t kill them and the exim queue run will not process those stuck messages due the spool file being locked by the stuck process.
I disabled TLS for the outgoing SMTP transport on one server and there are no more stuck processes.
> Aha. there was a recent for affecting FreeBSD which might
> be relevant: https://bugs.exim.org/show_bug.cgi?id=2831 <https://bugs.exim.org/show_bug.cgi?id=2831>
Actually this patch has been applied to the FreeBSD port when it was updated to 4.95. See:
https://cgit.freebsd.org/ports/commit/mail/exim?id=99c5dc1049a23570016dcb5ac44882e408800622 <
https://cgit.freebsd.org/ports/commit/mail/exim/files/debian?id=99c5dc1049a23570016dcb5ac44882e408800622>
It’s the mail/exim/files/debian/75_38-Convert-all-uses-of-select-to-poll.-Bug-2831.patch file.
I also added the following patches to the exim port and recompiled it:
BSD: fix resource leak:
https://git.exim.org/exim.git/commit/c57309a50444d858c0a2dc1581846a850d78a9ad <
https://git.exim.org/exim.git/commit/c57309a50444d858c0a2dc1581846a850d78a9ad>
Ensure server tls close alert not delayed:
https://git.exim.org/exim.git/commit/fc624b8cb4c3312d7450dfa86adfa3fe8dd9cbeb <
https://git.exim.org/exim.git/commit/fc624b8cb4c3312d7450dfa86adfa3fe8dd9cbeb>
I thought those patches have something to do with TLS and might fix the problem. But they didn't fix the issue.
I usually can find a few messages per day on a server so if you need me to test a patch, please let me know.
Thanks,
Michael
