-----BEGIN PGP SIGNED MESSAGE-----
I've been experimenting with mysql replication, specifically error
handling under varying error conditions. My tests are with 5.0.22 that
comes with CentOS 5.0.
My main concern is that I would like to lower the slave_net_timeout
setting, but don't want to risk the integrity of replication stability.
The default for the slave_net_timeout setting is 3600, which is 60
minutes. In my tests, setting it to 60 allows it to resume replication
when the master comes back up, some time less than 60 seconds, which is
So ... is slave_net_timeout = 60 a reasonable value to set on the
slaves? Google was unable to find anything other than a blurb that says
"consider reducing it", but that blurb doesn't give a guideline for
practical, real-world usage.
Setting it to 10 doesn't seem all that terrible either, though at that
point, any unforeseen severe spike in load resulting in mysql IO stalls
could cause the slaves to trip that timer and trigger a network
connection close and reopen. But I'm just speculating, I've not seen
this nor have I ever heard of this.
Why do I want to lower this setting? By lowering this setting to 60, I
can achieve roughly the same time to resume replication between a clean
shutdown and a hard (crowbar power) shutdown relative to when the mysql
master comes back up. It also removes a step in resuming replication of
"STOP SLAVE; START SLAVE" (which would kickstart the replication
process) because it hits the timeout rather quickly.
Of course this doesn't address potential errors with the Slave SQL
Thread, but that is known.
There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order. --Ed Howdershelt
Linux kernel 2.6.22-14-generic 7 users, load average: 0.02, 0.03, 0.00
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
-----END PGP SIGNATURE-----
|• Sane values for replication||Todd Lyons||6 Dec|