I'm not sure there's an easy way to speed that up. It depends a lot on your
configuration and your workload. As you said the slave needs to run the
same inserts/delete/update than the master but serially, there's some tools
(http://www.maatkit.org/doc/mk-slave-prefetch.html) that may pre load the
indexes to speed up some kind of queries, but it never did a good job to
me. You could also try to speed up your inserts/delete/updates queries
yourself making sure each query is hitting the proper index...
About server configuration, there's some thing you can disable on innodb to
speed up the queries but obviously it has it's negative things.
On Wed, Sep 26, 2012 at 2:58 PM, umapathi b <umapathi.b@stripped> wrote:
> no body there ?!
> - Umapathi.
> ---------- Forwarded message ----------
> From: umapathi b <umapathi.b@stripped>
> Date: Tue, Sep 25, 2012 at 10:16 PM
> Subject: cluster to non cluster replication issues ...
> To: cluster@stripped
> Cc: Johan Andersson <johan@stripped>, Jonas Oreland <
> jonas.oreland@stripped>, Wagner Bianchi <wagnerbianchijr@stripped>,
> suresh kumar <sureshkumarilu@stripped>
> Hi ,
> I recently re-started one of the data nodes of the production cluster to
> which the slave (innodb) was pointing .
> The node re-start almost took 3 hours . There was around 16 GB data in
> whole . During this time ofcourse , the slave was lagging behind .
> But after the re-start , the master has written two huge binlog files . The
> slave was just lagging behind for almost 3 days , executing all the
> in these files . This was not happened before like this , though I
> re-started the servers in order to free some memory earlier .
> Why this is happening ? I know that the slave has to read all the
> statements one by one as the SQL IO is single threaded .
> Still , I think there should be some way to avoid this . Could anybody
> guide me in this regard so that the slave never falls behind the master .
> Waiting for your valuable suggestions,
> - Umapathi.