On 15 May 2018 at 19:31, Chu, Raymond <[email protected]> wrote: We do use it extensively on our development environment to clean things up > after the evening compliance scanning and backups that are memory > intensive. It runs at a random minute in the 4am hour to avoid running it > on all at once. > > > > There are some thing that it doesn't get along with it: DB2, GPFS > (Spectrum Scale), and TSAMP. > > Clustering things don't like the "pause" so any of the Linux-HA stuff > would fall in that category. > > Ray, I have certainly seen it in use with DB2 and Oracle, for example. The more excess memory in the virtual machine, the longer it takes to give that back. The other virtual CPUs will also be held back intermittently, but nothing that wouldn't happen in a virtualized environment. If your applications are having time-out issues, maybe the threshold is too low. There is a cost involved with high frequency heartbeat messages, and that may not be always justified in a virtualized environment where a lot is covered already by the infrastructure.
I would not be surprised if the file system scanning itself would have a similar impact when you did not do cmmflush. The reason to do it all at once is to minimize the period with memory pressure. Since that is less of a concern with middleware that sits on the memory it once allocated, you might find value an approach that shaves off small chunks at a time (or not create virtual machines that are way too big). Rob > > > But works great for WAS servers and things like that. > > ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
