Which counters are interesting


The currently supported, clearcase specific counters are described here



We have deliberately not yet implemented measurements for mvfs - the caches - for 2 reasons, the major reason is that these caches depend on free memory so that is far more interesting to monitor, second all experiences with mvfs tuning shows that although their sometimes are measurable improvements to performance, the experience of that improvement is relatively insignificant, and it is heavily related to the usage scenario.

With respect to ClearCase performance the largest impact comes from the application layer, either from irrelevant 3'rd party applications running on the same server like continuous anti-virus scanning, databases or or anything else which is not required for running a ClearCase server.
Another wellknown performance thief is ClearCase trigger scripts that are too heavy, for instance by calling cleartool repeatedly - or any other tool implemented to enforce company policies.

At the very bottom, ClearCase runs on some piece of hardware in some kind of infra-structure, and you want to know if these parts are in optimal condition, so here are a couple of counters you can monitor

CPU

You'll never want the CPU('s) to max out - monitor for instance the counters:

\Processor(_Total)\% Privileged Time
\Processor(_Total)\% User Time

Memory

At any time you want to have available memory, swapping is bad, memory is way cheaper. Keep an eye on:

\Memory\Available MBytes

Disc

The disk system consists of the disk controller as well a physical - as wel as logical disk's

Monitor for instance

\LogicalDisk(C:)\% Free Space
\LogicalDisk(D:)\% Free Space
\PhysicalDisk(0 C:)\Disk Transfers/sec
\PhysicalDisk(1 D:)\Disk Transfers/sec
\PhysicalDisk(0 C:)\% Idle Time
\PhysicalDisk(1 D:)\% Idle Time


If at all possible keep the Operating system at one disk, eventually together with ClearCase application, and the data (vobs and views) on other disks.

More counters ?

Please send us a mail at info@praqma.net if you have suggestion for more ways of monitoring.