SAP on IBM i – Update week 30 2013: Kernel maintenance during uptime
Welcome to the SAP on IBM i update! You will have recognized that our team started a new effort to maintain this blog. By distributing the actual work over more members of the team we hope to reach better continuity.
This time I’d like to return to kernel maintenance. I know we talked about this several times already. But there is one detail, which I feel every admin needs to take into his or her considerations.
How it used to work…
What was a proper way to patch a kernel in the old days, when everything was built up from a kernel library? I suppose most of you did something similar to creating a copy of the library and applying a patch which APYR3FIX to this library. The library was a container that could be transported to any other system for or after testing. The actual update of the target system required downtime to run APYSAP (or its predecessor APYR3KRN) in order to create or update the corresponding file system objects. So preparation could be taken offline.
… and how it can work today
At latest with activating the downward compatible kernel 7.20 you faced a paradigm shift of how the objects in file system and in the kernel libraries are deployed. Deployment has to take place into the master directory DIR_CT_RUN. All convenience and safety actions of tool APYSIDKRN take place while the patch packages are examined. This includes saving off the existing kernel objects, checking for the right sequence of the packages that are deployed and many more and takes time – you probably do not want all this to happen while your system is down.
And here is my point: If your system is not configured for replication with sapcpe your instances are running on DIR_CT_RUN. All the described actions including potential errors in the process that need a fix will have to run during downtime. With replication, where every instance of an SAP system has its own copy of the kernel, this copy is updated automatically and quick at system start, while the expensive stuff can run offline for the master directory.
See Note 1632754 – IBM i: Changing to an instance-specific directory to get into the details.
Addendum July 30, 2013: Document SAP on IBM i Kernel Maintenance deals with the same topic