Greetings Fellow BBers,
My site uses the bb-iostat.sh script to monitor the virtual memory of
our our systems. It works fairly well on our Solaris systems, but the
defaults constantly generated errors for an excessive scan rate on
Linux. We found ourselves having to increase the scan rate from 3/6
to 500/1000, an alarming difference.
My first question is, "Has anyone else encountered this problem?"
I've done some digging and understand the issue much better. Linux is
not specifically handled by the script, so it drops to the default
handler when reading the output from vmstat. The problem is that
vmstat on Linux doesn't display a scan rate and the script grabs the
number of context switches instead. This can easily go into the
thousands or tens of thousands on multi-processor machines and has
little to do with the performance of the Virtual Memory Manager.
One can certainly increase the defaults with little worry because a
context switch takes CPU time which will be caught in other tests, but
it would be nice to have this fixed so there is less worry with future
users. I also note that the default check by the bb-iostat.sh script
doesn't check the system CPU usage, something that could be easily added.
I can make these changes, but I wouldn't want to step on any toes.
Craig Cook last modified the latest version back in 2002. Would there
be any problem is I made the changes and uploaded the new version to
deadcat?
Also, does anyone have any other thoughts on this?
Thanks,
Brian
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-==-=-=-=-=-=-=-=-=-=-=-=
To unsubscribe from this list, or to subscribe to the bb-digest list
send e-mail to mailto:majordomo@bb4.com with unsubscribe bb -and/or-
subscribe bb-digest in the BODY of the message.