On the other hand, InnoDB probably won't help. The replicated statements
are performed one at a time (to guarantee consistency), so they can never
fight with each other.
However, they can fight with reads; in this case InnoDB _may_ help.
You have an example of where slaves should be at least as powerful as their
Show us some of your common statements, together with SHOW CREATE TABLE and
SHOW TABLE STATUS.
> -----Original Message-----
> From: Boyd Hemphill [mailto:Boyd.Hemphill@stripped]
> Sent: Thursday, June 22, 2006 1:34 PM
> To: Frank; replication@stripped
> Subject: Re: Strategies to avoid slaves falling too behind
> Use RAID 10 or 01 on the slave will make your writes much
> faster. Since
> the slave is generally only written too in your case (I am
> assuming) you
> need all the through-put you can get.
> Since you are using MyISAM, disable the key writes (on the
> slave). YOu
> don't need the indexes.
> If you are reading from the slave, then that invalidates the second
> Best Regards,
> Boyd E. Hemphill
> Database Administrator
> Region XIII - Educational Service Center
> MySQL Certified Professional
> (512) 919-5413
> >>> Frank <softwareengineer99@stripped> 6/22/2006 10:54 am
> With my current replication setup during the peak period slaves start
> falling behind master and catch up once the peak period is over. At
> the slaves can be as much as 24000+ seconds behind. The master is
> twice as powerful as the slaves (8GB RAM vs 4GB RAM) which explains
> slaves cannot keep up during the peak period. I am currently
> slow queries to optimize but even then the slaves will not be able to
> up to date at peak times due to the difference in server specs.
> I am planning to convert the tables to use InnoDB as the
> storage engine
> reduce locking during the peak times.
> My question is that is there anything I am missing that can
> help slaves
> up to date without having to upgrade the hardware on the slave machine
> from optimizing queries? I am using MySQL 4.1 / RAID5 / MyISAM (don't