Understanding Polling and Metric Collection with Isilon

The Isilon Integration collects performance metrics from the Isilon cluster at the frequency of the time inter­val (e.g. 1 min) specified when configuring the integration. Performance metrics are requested with the spec­ified begin and end times for the corresponding time interval.

However, Dell EMC Isilon does not strictly observe the specified begin and end timestamps. It is therefore possible that the returned metrics data might have drifted by a few seconds.

The Isilon Integration addresses the time shift by normalizing the begin and end times (time-aligning to the nearest boundary) and storing the timestamps in the IO database. As an example, if you are col­lecting data at 10 second intervals, the data is aligned to the nearest 10 second boundary. Normalization offers the convenience of easier to read reports with negligible effect on the overall health monitoring.

When configuring polling, an interval of less than or equal to 5 minutes is recommended. This is because the low-level granularity of the Isilon metrics are only retained for a short period of time. Most metrics which offer live data are retained for 10 minutes, though some have even shorter retention policies. Therefore, if you gather metrics too slowly you will have gaps in your history due to Isilon aging out the data.

With more frequent metric collection intervals, IO collects and stores data for metrics at a lower level granularity. For example, if the metric polling interval is set at 5 minutes, then every 5 minutes all of the metric data will be collected and aggregated to the 10 second and 1 minute levels automatically, backfilling the data from the last poll cycle. However, a few metrics, such as the number of open files, will only be col­lected at the set poll interval (5 minutes in this example).