On 1/16/2013 1:01 PM, rob@xxxxxxxxx wrote:
Will a fully active system (about 900 interactive users and a few dozen
batch jobs) ended with
successfully flush it's cache?

Rob, probably not. But if you're plans are to bring the LPAR or system back up after doing some maintenance, it's not a problem. It's only if you're planning on rearranging disk or taking the system off the grid for days or weeks that ensuring the write cache is written is ultra important.

The delay on PWRDWNSYS, ENDSYS, ENDSBS tells the system how long to wait for jobs to recognize the system will be shutting down and do their own routines to end before it forces them to end.

You may increase the odds of all data getting from write cache to disk if you take down the system by first using ENDSBS *ALL or ENDSYS (both commands do about the same thing), then running PWRDWNSYS with DELAY(xx). A few seconds or a few minutes is probably not going to be enough time on systems that do a lot of writes. When you're ending the system a bunch of writes get generated (even if you say no joblog) because that's simply what happens when you end a job.

One more thing I just remembered .... there is a step in the power down process that instructs the IOAs to write out the data, an SRC gets posted when at this step. As I recall, you can sit on this SRC for up to 5 minutes while the hypervisor waits for acknowledgement from the IOAs.

This thread ...


Return to Archive home page | Return to MIDRANGE.COM home page