Processor utilization & Alert levels
nam
Posts: 6
Hi All,
I’m currently evaluating SQL Monitor and I’ve seen something strange.
With the configuration I have (It’s the default) the alert thresholds for processor utilization are at 85% for low, 90% for medium and 95% for high.
For one machine in a windows cluster, the Analysis is showing a “Machine: processor time†between 88% and 95% over a longer period of time. I was already wondering why an alert was not generated. (The delay is set at 60 seconds. But Analysis is not showing a drop below the 88% for several minutes)
Why isn’t an alert raised? Is Analysis not showing a drop below the 85% (status every 15 seconds) which is taken into account when not raising the Alert?
Eventually an Alert was raised: “Low Processor utilization†for that machine.
If you however look at the details?
Min. utilization: 92.7%
Max. utilization: 97.5%
Avg. utilization: 96.0%
So why, is this a low alert? And not a medium alert?
What am I missing?
With regards,
Niels.
I’m currently evaluating SQL Monitor and I’ve seen something strange.
With the configuration I have (It’s the default) the alert thresholds for processor utilization are at 85% for low, 90% for medium and 95% for high.
For one machine in a windows cluster, the Analysis is showing a “Machine: processor time†between 88% and 95% over a longer period of time. I was already wondering why an alert was not generated. (The delay is set at 60 seconds. But Analysis is not showing a drop below the 88% for several minutes)
Why isn’t an alert raised? Is Analysis not showing a drop below the 85% (status every 15 seconds) which is taken into account when not raising the Alert?
Eventually an Alert was raised: “Low Processor utilization†for that machine.
If you however look at the details?
Min. utilization: 92.7%
Max. utilization: 97.5%
Avg. utilization: 96.0%
So why, is this a low alert? And not a medium alert?
What am I missing?
With regards,
Niels.
Comments
I've just quickly tested this alert several times and the severity was as expected.
The values on the details tab can change if the alert changes severity during the lifetime of the alert so it's possible that the alert was initial raised as Low but was then escalated to Medium. It's possible to see if this happened by going to the Alert History tab on the alert details page.
It's also worth checking the "Machine: processor time (%)" graph on the Host machine tab in the Performance data section of the alert details. This has two vertical lines (grey / green) that indicate when the alert was raised / ended. Do these still suggest that that the alert should have been Medium to start with?
Regards
Chris
Test Engineer
Red Gate
Unfortunately not. History is only showing the Low being raised and being ended. (1 minute apart).
Yes, but unfortunately I found out that this graph is not accurate enough.
Looking in the Analysis page I can see that one minute before the alert was raised the value was 89.9% So, the alert level low is correct.
After that the value was always above 90.0% (Actually 95.0 until the alert was raised).
The alert was ended one minute after being started. The value at that time was 85.2%
I found out that de information shown in the detail was from the moment the alert was raised until (but not including) the moment the alert is ended.
The question I have is, should 85.2% be a value on which the alert is ended?
Even if we exclude the value 85% as an "alert value". In the 6 minutes before the alert was raised the value (15 second interval) was never below 86%. During that time, there actually was a period of 2,5 minute that the value was not below 91%. For some reason no Alert was raised.
I gathered the information with Analysis and using an export.