Performance Counter

[hmo]

In general:

Establishing a Performance Counter seems to be very slow. Following tests seems to stall HostMonitor with different results. Sometimes it's "bad" and "Unknown".

A simple test like this:

10.151.128.50Memory% Committed Bytes In Use

or

10.151.128.50MemoryAvailable MBytes

can take up to 2-3 min. just to make.. and eventually the server will repond to HostMonitor. Maybe 1-2 tests out of 10 will be "bad" or "Unknown".

HostMonitor is located on the same C-class - so no slow router/dsl-connections to take care off... on the other hand, if I make a Performance Counter to a computer sitting on a 256Kbit DSL-line (transparent net - no firewall what so ever), it can take for ever... (and NO... the dsl-line is not flooded with any trafic )

Process tests seems to have "some of" the above "illness" as well....

Hints: HostMonitor is running 3.64 on XP and monitored server is W2K standard (up to level with SPs).

Any clue?

Ohhh.. found this in Your doc.:

******
Important notes:
Note #1 (permissions): The security on the following registry key dictates which users or groups can gain access to the performance data: HKEY_LOCAL_MACHINESOFTWAREMicrosoftWindows_NTCurrentVersionPerflib. In order for to query performance data, account must have KEY_READ access to the above registry key. To change permissions to the registry key you can use the registry editor utility (Regedt32.exe).

Note #2 (bugs): Windows

******

I can't figure this out!


[ This Message was edited by: hmo on 2003-02-16 15:00 ]

[ This Message was edited by: hmo on 2003-02-16 15:01 ]

[KS-Soft]

I am afraid there is nothing we can do HostMonitor uses Windows API to perform this test. If Windows needs 3 min to retrieve information from remote system, we cannot change this. Om some systems Performance Counters work good, on some systems it very unreliable. Why? Microsoft only knows. We did not find any explanation on msdn.microsoft.com.

Regards
Alex

[hmo]

Hmmm...

Untill now I have'nt seen any reliable systems within our +450 servers at all! It might be Microsoft to blaim? On the other hand... the use of Performance Counter(s) among others IS very important. This might be important to implement into a new version (ie. working through remote agents). In that case You can attack Performance Counters locally- and even without use of MS API!

I did some (in fact - many) tests on W2K and XP (local and remote by dsl connections). To semilar tests against the same computer can come up with different results... mostly on of them did a "bad/unknown" and the other one was okay! How come that the same test can figure out to "burn" one of them?

You cant do anything?

Cheers,
Hans Mosegaard

[KS-Soft]

Both "Ok" and "Bad" statuses mean Windows API successfully retrieved information. Why status sometimes "Bad" sometimes "Ok"? I don't know. What exactly test settings do you use (you can use Export function to get text file with test settings)?

Regards
Alex

[dimmyr]

Alex, you do not give directions to the Performace Monitor Test setup. You mention in the Help that you need to give KEY_READ to the PerfLib key, using the registry, but how do you do that?

Note #1 (permissions): The security on the following registry key dictates which users or groups can gain access to the performance data: HKEY_LOCAL_MACHINESOFTWAREMicrosoftWindows_NTCurrentVersionPerflib. In order for to query performance data, account must have KEY_READ access to the above registry key. To change permissions to the registry key you can use the registry editor utility (Regedt32.exe).

[KS-Soft]

>You mention in the Help that you need to give KEY_READ to the PerfLib key, using the registry, but how do you do that?

By default administrator has this permission. If you want to change it,
- start regedt32.exe,
- select HKEY_LOCAL_MACHINESOFTWAREMicrosoftWindows_NTCurrentVersionPerflib key,
- click Menu->Security->Permissions

Regards
Alex