On 01/11/2019 10:16, Reio Remma via dovecot wrote:
On 01/11/2019 01:19, Benjamin Connelly via dovecot wrote:
during the update the log_warnings changed from 1 to 2 therefore showing lots of aborted connection notices in the logs
changing the log_warnings back from 2 to 1 solved this issue
Yes the same setting made the same change to the default with mysql:
https://dev.mysql.com/doc/refman/5.7/en/server-options.html#option_mysqld_lo...
So now we know how to silence the "Aborted connection" mysql loglines thank you!
But still I wonder if Dovecot wants to handle these connections slightly differently - if it could be cleaner? Or is it moot?
Unfortunately there are (replication info) messages that I actually need from that log level and I wouldn't want to just hide issues anyway. :)
I monitored the situation in MySQL Workbench a little and it seems the userdb MySQL connection closes cleanly.
What doesn't close cleanly is the dict engine MySQL connection that is updating last login timestamp (its connection shows COMMIT as the last query).
The dict connection closes after sleeping exactly 60 seconds (server net_write_timeout is 60 seconds). The userdb connection closes after about 61-62 seconds for some reason.
I tried changing read/write_timeout in Dovecot MySQL connection to lower, equal and higher than those set by the server, but it didn't change anything.