What are top 5 counters to monitor from Windows performance counters except disk and memory usage?
It depends on the services running on a node. These would be different on an interface / PI Data Archive / Coresight, etc.
KB00884 - PI System daily health check
KB00830 - Monitoring interfaces with health points
KB00047 - Interface Status Utility (ISU) and UniInt Health Points: What are the benefits of each
KB00597 - PI ACE performance monitoring
they all good but i was asking about Wondows only not PI.
Top 4 windows counters in adition to disk and memeory usage (I already have them on my ourdashboard).
In general my approach is that if a server is there to run PI, the performance of PI should be monitored, and to a lesser extent the performance of the underlying Windows host.
But a few key things that i normally monitor are disk space levels (<20%), CPU levels (>80% for >5 minutes), Available physical memory (<20% for >5 minutes).
It really depends on the service levels to deliver, system sizing, system purpose, design, , virtualisation, maintenance, etc. Depending on those you could add counters to monitor specific items.
How can I setup this trash hold, memory < 20% for > 5 min,
I can set memory<20% but how to make for >5 min, I don't know.
What i do in general is to create an AF model for the PI system components. For e.g. the node hosting the PI Data Archive is have an overall CPU usage PI Point. In a separate AF attribute i calculate the 5 minute rolling average using time-range overide, relative time of *-5m and average. In a third attribute i set the threshold, e.g. 80%, and in a fourth i calculate a true/false trigger if the 5 minute average exceeds the threshold. I use the latter to trigger EventFrames / Notifications.
it quite doesn't make sense from answer, I would appreciate if you put screenshots how to do it in steps.
Are you using PI Notifications to trigger an alarm based off of this low memory threshold? If so, then we can set a "Time True" value within the Notification itself (for Notifications 2012)
This means that the Notification will identify when your Input counter (in this case, Available MBytes) becomes less than our threshold (here, 1000), but it will not send the alert unless the value remains below that threshold for the length of time specified in our 'Time True' parameter
Retrieving data ...