Loading Data into New Cluster Woes

[Terry Lillie]

I'm attempting to run a performance benchmark on a simple cluster configuration.

My test environment consists of two 3.2 GHz, 1 Gbyte systems, both running SuSE Linux 9.3.
One machine is configured to run mysqld and the other to run a single ndb data instance and
a single ndb manager node. I'm running MySQL MAX 4.1.12 on both nodes.

Ultimately, I would use two separate systems for the data nodes, but for the moment, I'm just
trying to assess the performance advantage of using the cluster.

I have an existing database that consists of approximately 30 MyISAM tables that I'm trying
to convert use the cluster environment. A few of these tables contain 300,000 to 400,000 rows.

When I tried to convert the existing MyISAM tables using "ALTER TABLE X ENGINE=NDBCLUSTER",
I would see several "Got temporary error 4010 'Node failure caused abort of transaction' from
ndbcluster" error messages for the large tables. The smaller tables converted without incident.

When I look in the manager log file, I see this whenever this error occurs:

2005-06-29 14:42:59 [MgmSrvr] ALERT -- Node 2: Node 3 Disconnected
2005-06-29 14:42:59 [MgmSrvr] INFO -- Node 2: Communication to Node 3 closed
2005-06-29 14:43:02 [MgmSrvr] INFO -- Node 2: Communication to Node 3 opened
2005-06-29 14:43:02 [MgmSrvr] INFO -- Node 2: Node 3 Connected
2005-06-29 14:43:02 [MgmSrvr] INFO -- Node 2: Node 3: API version 4.1.12

In and attempt to address this, I wrote a PERL program that dumps the large tables (in
mysqldump format) into chunks of 1000 inserts and then imports these into the cluster --
pausing 4 seconds between each load of 1000 records. The bottom line is, I'm still
periodically seeing the same error message.

Is anyone else seeing this, or does anyone have any suggestions on what I might try?

Terry Lillie
OMIX, Inc.