Greetings,
On Wed, 28 Feb 2024 22:15:55 +0100, Kirill A. Korinsky wrote:
As the next step I've reduced logs to verbose, let's see how it goes.
Reducing verbosity re-triggered an issue.
Meanwhile, as a lucky guess I've increased the number of open files on all users which is used by dovecot as:
mx1$ cat /etc/login.conf.d/*
_dovecot:
:openfiles=1024:
:tc=daemon:
dsync:
:openfiles=1024:
:tc=daemon:
vmail:
:openfiles=1024:
:tc=daemon:
mx1$
My user has a lot of IMAP folders, and if I run doveadm fetch from shell with default limits (open files 128 files) it puts a lot of errors like:
doveadm(kirill@korins.ky): Error: fetch(guid) failed for box=virtual.All uid=5097: Internal error occurred. Refer to server log for more information. [2024-04-18 14:39:15] (for backend mailbox Archive.Texunatech) doveadm(kirill@korins.ky): Error: Mailbox Archive.Texunatech: open(/var/vmail/kirill@korins.ky/Maildir/.SomeFolder/dovecot-uidlist) failed: Too many open files
So, somewhere here I made a guess errors are swallow, but it exists.
So, I've moved future by this way.
- Shutdown smtpd and dovect.
- Disabled smtpd from autostart.
- Disabled replication on both servers.
- Update logins.conf to increase limits.
- Remove all virtual folders from all users.
- Reboot server.
- Wait couple of minuts (like 10) and enable replication.
- Wait one full sync cycle.
- Enable smtpd and start it.
After this I do have consisten output:
mx1# doveadm fetch "uid guid" -u kirill@korins.ky mailbox virtual.All | md5 8978c5e506bf7bcf769232d48b964e00 mx1#
and
mx2# doveadm fetch "uid guid" -u kirill@korins.ky mailbox virtual.All | md5 8978c5e506bf7bcf769232d48b964e00 mx2#
Anyway, that setup works for a day or two and I can't say that it really help because I need to update machines to OpenBSD 7.5 to update OpenSMTPD to fix another issue. This update included install of dovecot to 2.3.21 (47349e2482).
So, if issue do not reappear in couple of weeks, I can't say that help: increase of limits or update of dovecot.
Anyway, I'll be back.
-- wbr, Kirill