Automating iCluster Start-up – Part 1
Let me ask a pretty simple question – how often do you IPL? In the old days most shops religiously did IPLs on a weekly basis along with perhaps a full system save, and this was usually done to clear up address spaces by returning unused addresses to the OS, along with applying delayed PTFs etc. However with the incredible power of Power Systems running IBM i today, a weekly IPL is overkill. An IPL once a month seems to be the prevailing accepted frequency. When you IPL, does iCluster come back up or do you manually have to start the subsystem, nodes and groups?
Most customers have modified their start-up program specified in the QSTRUPPGM system value to start the XDMCLUSTER subsystem, then start the nodes and finally start replication groups. Prior to iCluster 7.1 TR2 this usually required a number of delay job commands in the CL to wait for TCP/IP to come up, wait for the subsystem and nodes to start, and manually specifying which groups to start. At iCluster 7.1 TR2 we have made this much easier for you. Two new commands DMSTRCST and DMENDCST have been added to iCluster to help automated start-up and shutdown. If you used the HSTRCST and HENDCST commands from DMXtras, you will recognize these two new commands as versions of those older commands.
The DMSTRCST command will be covered in the next post, but for it to do its magic requires a new group parameter called AUTOSTART, the very last parameter on the DMADDGRP command. Set this value to *YES for any group that you would like to automatically start via the DMSTRCST command. Refresh only groups will never be automatically started regardless of the value of this parameter.
In the next post we will cover the DMSTRCST command in its entirety.