Avg Meaning

[blissful holly]

Hi All

I have one question - when I do check NAbox provides grafana I see their is Average read and write latency , how do we read this , this is average of what and for how many minutes, thank you

[pliant ferry]

@blissful holly By default, Harvest collects performance metrics every 1 minute, so any calculation around performance metrics is for a 1-minute interval. The metric calculation document is available here (https://netapp.github.io/harvest/latest/configure-zapi/#metrics_1).

[blissful holly]

Thank you Rahul I wil go through this

[blissful holly]

Hey Rahul I have one question , sorry if it looks like stupid question- why in every map like Node Write latency we see value as Average, if I take example of top 5 nodes by write latency then its shows As nodes vs Avg , I’m still unable to understand, can you please help me on this

[hidden void]

@blissful holly,
I believe you would be referring this Avg column in right side of the table. So, as per the query used in this panel, it's node_read_latency and based on the variables from dropdown and topk, it would be actually displays instant(current) value of that Node (left column in table).

Here, This is performance counter and it's directly not available from Ontap and Harvest need to cook as per the Ontap performance calculation mentioned here: https://netapp.github.io/harvest/25.05/configure-zapi/#metrics_1, This xxxxx_latency counter should be having property as average.

You can find the Ontap description of this counter in our doc here: https://netapp.github.io/harvest/25.05/ontap-metrics/#node_read_latency, which clearly says this is average value based on the time interval(which is 1m by-default for perf)

[blissful holly]

Hi @hidden void thank you so much and apologies for late reply as I was unhealthily , As I understand you are saying the Netapp Takes 1 minute value and call this as average. So this value is of average 1 minute but graph shows real time value , please correct me

How it’s calculated: Harvest takes two consecutive measurements (polls) of the read operation latency from ONTAP, typically over a 1-minute interval by default. It then calculates the average by dividing the total latency by the number of read operations (based on a “base counter” like read_ops).
• Why it’s an average: Instead of showing the latency for a single read operation, it gives a smoothed-out value over the time interval (e.g., 1 minute), which helps understand the typical performance rather than momentary spikes or dips.

[blissful holly]

@pliant ferry and @hidden void it’s a request if you can please help me on this thank you

[empty merlin]

Hi @blissful holly most ONTAP performance metrics are monotonically increasing and provided in a raw format - Harvest cooks them based on the ONTAP metric type. Here's an example. You can query your cluster for the same if you want curl --silent --insecure -u"$user:$pass" 'https://$ip/api/cluster/counter/tables/volume' | less

The volume read_latency metric is of type average which means to convert the raw values ONTAP provides into a cooked value requires two polls and the applying the formula for "average" in the pasted table. By default, the two polls will be one minute apart.

Harvest then exports those cooked metrics and Prometheus scrapes them.