Some DB2 commands for the pureScale feature that saved my box...

DB2 nanoCluster

Some days ago I told you about db2greg and how I used it to repair a DB2 pureScale demo cluster. Well, there are more commands that help to repair an inconsistent cluster, some are new DB2 commands. I will describe them in this article (so I can quickly find them later on).

Before I start let me tell you about the demo cluster which is nicknamed "DB2 nanoCluster". It is inexpensive hardware like Intel Atom CPUs, 4 GB of RAM and Gigabit Ethernet. The system uses SUSE Linux and DB2 9.8 (which is the pureScale feature). It consists of 3 machines, one used as storage node and two others hosting a DB2 member and a cluster caching facility each. It can be built for few hundred Euros or US Dollars and fits nicely into a box or between shelves (see the picture). Because of the (insufficient) hardware and the difference to a production system, there are also performance differences and, important, missing redundancy. If you pull 2 power cables, including the one on the storage box, expect to have problems. Pull an Ethernet cable, have fun as a single cable would be redundant connectivity to the storage server in a production system, multiple Infiniband adapters, and external network to the application cluster. Anyway, the nanoCluster and DB2 work nonetheless - at least in most cases, hence the repair commands...

What commands are useful to repair inconsistent configuration?

"db2iupdt -fixtopology" comes in handy when adding or removing cluster components resulted in a timeout and inconsistent state (someone pulling a cable by mistake?). It is used to repair the instance configuration.

"db2cluster -cm -repair -resources" can be used to fix inconsistencies between the db2nodes.cfg file and the cluster manager resource model (the RSCT layer).

Sometimes there are alerts by the DB2 members that need to be taken care of. "db2cluster -cm -list -alert" displays them, "db2cluster -cm -clear -alert" clears them.

The "db2instance -list" command shows the current status of the entire cluster.

That's it for today, my cluster is up and running. Do you want to build and operate your own DB2 cluster? Take a look at this page.