I had never before used the Event Scheduler with Cluster (only with
InnoDB on a single machine). Now that I began to experiment with it on
Cluster it dawned on me that mysql.event is a MyISAM table and
therefore won't be replicated in the Cluster.

My goal is to store Events in the Cluster itself, so that they will be
run even if one of the SQL nodes is down.

I ran "alter table event engine=ndb" on both SQL nodes and that seemed
to work: Events propagated as expected and the recurring Event was run
(once) each time, as expected.

However, after stopping the cluster and trying to start it again I get
errors when trying to start the mysqld:
101209 17:48:01 [Note] table './mysql/event' opened read only
101209 17:48:01 [ERROR] Event Scheduler: Failed to open table
mysql.event
101209 17:48:01 [ERROR] Event Scheduler: Error while loading from
disk.
101209 17:48:01 [Note] Event Scheduler: Purging the queue. 0 events

I ended up deleting the NDB version of mysql.event and recreating the
MyISAM version :-( :-(

What's the best practice for using the Event Scheduler with Cluster?
I couldn't find Cluster-specific notes on the Event Scheduler.

Thanks....

Theodore Hope <theodorehope (AT) gmail (DOT) com> wrote:

More to that: the event scheduler thread runs in the SQL node.
And while you *could* activate the event scheduler on multiple SQL
nodes, it will not be synchronized. So where's the point to store
the events table in the cluster anyway?

This won't work with the current implementation.

Chose one SQL node to run events. And find a way to cope with it
when this node goes down.


XL