We have a mail server running Dovecot (2.3.x) and postfix (3.5.x) on RHEL8, using the RPMs from RedHat. Postfix is configured to use the Dovecot lmtp for deliveries to the local maildir for each user.
This server uses a virtual user setup, all mail goes under /home/vmail/<domain>/<user>/Maildir and the virtual_alias_maps are all LDAP based. We've had a few incidents where we have a large blast of email that comes in - usually caused by multiple services and systems all reporting a network outage or some other large event. When that happens, the mail server queue gets backed up, the load on the server gets extremely high (30-40), and these errors pop up in the logs: postfix/qmgr[ZZZZZZ]: warning: mail for foo.bar.com is using up X of Y active queue entries where X > Y e.g 6040 of 3065 and 'foo.bar.com' is the mail server itself. The load increase based on watching htop seems to be related to a large number of postfix and dovecot processes being spawned to process this flood of email. The queue gets backed up and I have to use a script to run 'postqueue -i' for each queue entry to force delivery, otherwise the email sits in the queue for hours as the mail server churns away. I've read through the documentation and what I can find online regarding that error, but I'm not sure what to adjust. This host is a virtual machine and I've increased the number of CPUs and increased the host memory, that helped some but this feels like some default value I have needs to be changed. Is this a case where I should increase the 'lmtp_destination_concurrency_limit' which is currently set to the 'default_destination_concurrency_limit' of 20 ? Or are there other parameters I should look at ? What factors determine the size of the number of 'active queue' entries ? Any guidance or pointers are greatly appreciated. _______________________________________________ Postfix-users mailing list -- postfix-users@postfix.org To unsubscribe send an email to postfix-users-le...@postfix.org
