Reply to this email directly or view it on GitHub.
Ok, so I tested your code in a production system and just replacing the module didn't always seem to do the trick, I still ended up with messages like these:
Nov 25 15:35:26 /usr/sbin/kamailio[15053]: ERROR: db_mysql [km_dbase.c:128]: db_mysql_submit_query(): driver error on query: SSL connection error: socket layer receive error (2026)
Nov 25 15:35:26 /usr/sbin/kamailio[15053]: ERROR: <core> [db_query.c:235]: db_do_insert_cmd(): error while submitting query
The error above lead me to the conclusion that "mysql_ping()" wasn't performed on these requests. Dug into the MySQL C API and from what I understand the MYSQL_OPT_RECONNECT=true is only used by and acted upon by mysql_ping(), no query functions. So I decreased the "ping_interval" setting to:
modparam("db_mysql", "ping_interval", 5);
to force mysql_ping to run more frequently. This code has now been running for almost 24h on a production system and it looks promising. Made a small addition to your code to get a warning if connection ID changed after mysql_ping (see attached diff). Now I'm getting a few of these messages:
Nov 26 08:55:27 /usr/sbin/kamailio[29363]: WARNING: db_mysql [km_dbase.c:91]: db_mysql_submit_query(): mysql thread id changed due to reconnect in mysql_ping
but from what I can see no more failed INSERTs.
I'm thinking MYSQL_OPT_RECONNECT together with frequent mysql_ping()'s seems to have done the trick.
—
Reply to this email directly or view it on GitHub.