MDEV-36002: --master-retry-count=0 can still stop #4220
Conversation
If master-retry-count equals 0, the slave io will try to reconnect to the master after sleeping master-connect-retry seconds. In this case, the instructions to update reconnection counter(err_count) won't be executed, thus avoiding the err_count overflow issuse.
|
|
There was a problem hiding this comment.
@andrelkin gave an excellent idea on testing the overflow wraparound on the Zulip thread.
There was a problem hiding this comment.
Please include the default my.cnf file (in the rpl folder, IIRC) instead.
Then you’ll only need the master-retry-count line.
Actually, can you include a non-zero master-retry-count through a .combinations file instead?
In a .combinations file, sections specify different configs. Configs apply to all servers (both primary and replica), though the primary doesn’t care about this case.
In either case, test names should be descriptive, if available; i.e., rpl_master_retry_count.
Include a line break at the end of the last line (where GitHub labels ⛔); thanks.
There was a problem hiding this comment.
Ok, I will do it and thx for your advice. This may take some time though, as i am still getting familiar with mariadb content.
Co-authored-by: ParadoxV5 <paradox.ver5@gmail.com>
If master-retry-count equals 0, the slave io will try to reconnect to the master after sleeping master-connect-retry seconds. In this case, the instructions to update reconnection counter(err_count) won't be executed, thus avoiding the err_count overflow issuse.
Description
the patch fix the bug that when --master-retry-count=0 the reconnection can still stop for an overflow issue.
Release Notes
No update, just as kb says "A value of 0 means the replica will not stop attempting to reconnect."
How can this PR be tested?
master_retry_count is set as 0 and master_connect_retry as 1 second. The master will be shutdown for 5 seconds, during this time the replica should attempt to reconnect to the master. After 5 seconds, the master will be restarted, and the reconnection from the replica should success, which means it's Slave_IO_Running should be Yes.
Basing the PR against the correct MariaDB version
mainbranch.PR quality check