Master hanging up on replication?
I am currently experiencing a problem with a server located in a dedicated hosting company. The machine is all mine, and runs Debian Woody, Apache, mod_perl and MySQL 4.0.21 (compiled from source). The slave is my home workstation, on a cable modem. It runs Slackware 10.1, with the same version of mysql, compiled from source. This setup was working fine for months, but last week the server started to die without warning or explanation - it would just "go away" and I couldn't connect on any port. Reboot was the only way to recover, but then it would go down again randomly, anywhere from a few hours to a day later. There were no errors in the kernel logs or mysql logs. I suspected a hardware fault, and the hosting company first swapped the RAM, and then the whole server hardware (i.e. moved the hard drives to another box). But, it still happens. So, I began to suspect software, and I seemed to remember in the past, I had seen the server hang when I had tried to use a certain version of mysql on the client side as a slave. Somehow the slave was able to totally crash the master, but I couldn't remember the details. I think it maybe had something to do with using RedHat 7.3, which had a broken version of gcc. Anyway, I tried turning off replication on the slave at home, and now the server seems to be staying up.
Has anyone seen this kind of problem, with a slave being able to crash the master - not just the mysql process, but the whole server?
Could this have anything to do with the master being compiled with an earlier version of gcc (Debian Woody - gcc 3.3.2, compiled from source) than the slave (gcc 3.3.4 on Slackware)?
Any experiences welcomed. I haven't upgraded to 4.1 yet because my applications use TIMESTAMP in a certain way and that format seemed to have changed between 4.0 and 4.1.
Thanks,
-Neil