Subscribe

Issue with PA counters -vs- ONTAP CM counters..

I have a situation at a customer were that ONTAP CM counters for volume latency do not match at all, even if you tried to account for the hourly-average nature of the CM counters.

The CM Counter view:

Attached

The PA Counter view:

Attached

They just dont match each other.  The CM counter chart is showing a higher average latency than the PA chart.

My question is:

Is there a possibility that there is separate system (low priority) volume IO that is raising the average volume latency seen in the ONTAP CM data, that we are not seeing in the PA average latency data?

Issue with PA counters -vs- ONTAP CM counters..

I think they look almost similar. Here is why I feel so,

In PA graph, the brown line with higher values could be other_latency and the avg_latency brown line overlaps with the read_latency line. ( this can be confirmed by unchecking the other latency and read_latency lines). There could be a confusion on this becuase of the color of the avg_latency and other_latency lines are almost same.

In CM graph, avg_latency lines shows the value from 8K-10K ( I am considering the right axis as all the three counters are measured in terms of microsec and easy to compare them with PA graph ) which is in line with what we see in PA.

Please correct me if I am wrong.