Any opinion on this otto?

Can you rephrase the question as a diff? To me https://anonscm.debian.org/cgit/pkg-mysql/mariadb-10.0.git/tree/debian/mariadb-server-10.0.mysql-server.logrotate and https://github.com/MariaDB/server/blob/10.2/debian/mariadb-server-10.2.mysql-server.logrotate looks good.

Once implemented in mysqladmin would look something like:

diff --git a/debian/mariadb-server-10.2.mysql-server.logrotate b/debian/mariadb-server-10.2.mysql-server.logrotate

index 2a9c3cd..c97217b 100644

--- a/debian/mariadb-server-10.2.mysql-server.logrotate

+++ b/debian/mariadb-server-10.2.mysql-server.logrotate

@@ -13,7 +13,7 @@

           test -x /usr/bin/mysqladmin || exit 0

           if [ -f `my_print_defaults --mysqld | grep -oP "pid-file=\K[^$]+"` ]; then

             # If this fails, check debian.conf!

-            mysqladmin --defaults-file=/etc/mysql/debian.cnf flush-logs

+            mysqladmin --defaults-file=/etc/mysql/debian.cnf --local flush-logs=error,engine,general

           fi

        endscript

 }

Its intentionally not instructing mariadb to rotate the binary logs at the log rotate time. This because of reliability issues (that I'm sure will be worked though) but also this has an effect on large transactions that may be occurring at log rotation time. Binary log are controlled by max_binlog_size and expire_log_days so the lack of a daily/weekly logrotate seems like a good tradeoff.

LGTM

To be clear, the server side of this implementation by support of various individual FLUSH options (https://mariadb.com/kb/en/mariadb/flush/) is already complete. This just requires the client mysqladmin to implement mapping command line options (as suggested above) to the SQL.

serg, I'm hoping this isn't too late for 10.0.30 and in light of the large incomplete list of crash bugs references here who suffer segfaults as a result of flush-logs this was an easy fix. All it does is stop flushing binary logs during logrotate.

Only mysqladmin and associated scripts are changed.

No server core changed in the attached PR.

danblack, sure, I think it's a good change even unrelated to other bugs.

But I don't understand this --local flag. Where did you find mysqladmin that supports it?

It didn't support it. I added it. https://github.com/MariaDB/server/pull/323/commits/be82664a331808637755448948c8be7ba14aa258