I Have 8 windows servers in a network being monitored by a remote probe on server 2008 machine.
Most servers are 2003 standard (NOT R2) All servers were added to PRTG by using the "STANDARD NETWORK DISCOVERY". Network discovery nicely populated WMI sensors for disk/memory/network,etc as per normal.
WMI monitoring ran great for about a week, then on 4 of 7 server 2003 machines, i started getting "WMI Query returned an unusual result - try alternative WMI settings". The failing servers host the following applications: server1-MS SQL & HTTP interface, server2 Sybase, server 3 CITRIX, server 4 some 3rd party data interchange programs.
During the workday, these machines run average CPU utilization of 5-8%, peak around 20-30%. RAM utilization peaks around 85% at times during the day, but no excessive paging - nor substantial virtual memory usage (teven though hey are 5 years old!)
I changed the WMI config on these machines to alternate WMI settings as recommended and now am getting PE054 WMi timeout messages.
I cycled the WMI service on the MS SQL server and about 1/2 of the sensors were getting data wheni started typing this message, the other half were not. I changed the WMI update interval to 180 seconds because i saw a log entry WMI update last result was 74%, over 50% is a timeout. I just checked that machine again and now all but the PING and HTTP sensors are WMI TIMEOUT
Given that other mahcines in this network that are the same vintage, with same load characteristics are running fine I'm wondering what the PE054 is really saying - does that mean that the remote probe is being overwhelmed by all the sensors reporting in, or am I correct in interpreting this result as the probed device sensor failed to reply withing a speficied amount of time?
In this environment, would i be better off using SNMP based sensors rather than relying on microsoft?. In delivering EMI, have they really provided yet another buggy, crash prone, bloated, unreliable and erratic networking fundamental (did anyone ever succeed trying to actually use WINS instead of DNS?)
At least with SNMP i may have a chance of getting a consistent result.
It's very frustrating to have bought 1000 sensors only to find that when the mahcine is moderately busy I can't monitor iti!!!