Myslqd

[manthrax3]

Every so often mysqld pins my server's processor (100%) for about 1-2 minutes. I'm not sure of the exact frequency, but it's high enough that I've noticed it in top.

This is a a pretty decent server: AMD Sempron(tm) Processor 2600+ with 1G of ram.

Ideas?

[KevinH]

What version do you have? M2 had a new logger process that didn't have enough data indexes. This will be fixed going forward.

[manthrax3]

Originally Posted by KevinH

What version do you have? M2 had a new logger process that didn't have enough data indexes. This will be fixed going forward.

I'm using M2. Is there a sql script I could use to update the log tables to add these indexes? This is causing some serious performance problems for me.

Thanks!

[marcmac]

su - zimbra
logmysql
use zimbra_logger;

alter table mta add index i_arrive_time (arrive_time);
alter table amavis add index i_arrive_time (arrive_time);
alter table mta_aggregate add index i_period_start (period_start);
alter table mta_aggregate add index i_period_end (period_end);
alter table amavis_aggregate add index i_period_start (period_start);
alter table amavis_aggregate add index i_period_end (period_end);

[manthrax3]

Originally Posted by marcmac

su - zimbra
logmysql
use zimbra_logger;

alter table mta add index i_arrive_time (arrive_time);
alter table amavis add index i_arrive_time (arrive_time);
alter table mta_aggregate add index i_period_start (period_start);
alter table mta_aggregate add index i_period_end (period_end);
alter table amavis_aggregate add index i_period_start (period_start);
alter table amavis_aggregate add index i_period_end (period_end);

Yeah, that fixed it. It used to take 2 minutes at 100% processor. Now it takes 3 seconds at 7%.

Thanks for the quick turnaround.

bp