We've been running Zimbra open source for a few years with about 10k accounts, yesterday afternoon MySQL started refusing connections it appears. I've seen too many connections, and the 75%, I raised the values substantially and it doesn't seem to make any difference.

So far, I have rebuilt the tables, changed the memory, changed the hardware, changed the drives, and I still keep getting oom-killers in dmesg, which I believe are caused by the mysql issue. If I netstat -np|grep 7306|wc I can see the count go up to about 280 or so and stay pretty much there, until it crashes. Mysql_error.log doesn't show any crash reason just the startup notice. I'm at a loss at this point as to what could be causing it to crash. It can crash in 15 minutes or an hour, I can't find a pattern. Any pointers would be immensely helpful.

Should the connections to the MySQL server linger like that?