Hello, I am trying to replicate performance metrics from active IQ unified manager into harvest. Major difference when comparing volume latency is that AIQUM uses ms/op instead of ms. I wanted to create this same metric, but I am having a tough time making the metric work. My understanding is that we need to divide volume_avg_latency by total_ops, is this the correct approach?
#Possible to convert ms to Ms/op?
1 messages · Page 1 of 1 (latest)
@swift tide Harvest volume latency is per ops only. Harvest calculates its latency values by using its base counter ops.
https://netapp.github.io/harvest/25.05/ontap-metrics/#volume_avg_latency
Thanks @rustic thorn , I am running harvest 25.02.0 btw. Not sure if that makes a difference. Regardless, we are seeing pretty large discrepancies between the latency via harvest and AIQUM during the same time frames. Now I am wondering if the averages across a timeframe are different on each platform. I can’t send a screenshot as my org blocks discord, but I would be happy to email my findings?
@swift tide It's best to compare the latencies provided by ONTAP with those from Harvest. Try running the ONTAP CLI command below in diagnostic mode to obtain the volume average latency for comparison against Harvest.
statistics show-periodic -interval 60 -iterations 50 -object volume -instance osc_vol01 -counter avg_latency
Harvest also has workload level latency for volume. You may check that as well.
https://netapp.github.io/harvest/25.05/ontap-metrics/#qos_latency
25.02 version is fine.
Interesting. The statistics command seems to be aligned well with what harvest is showing. Ive noticed AIQUM collects at a 5 min interval, whereas my harvest instance collects at 2 minutes. I wonder if that has anything to do with the difference in latency graphs as well
Collection interval shouldn’t change values drastically. See if harvest metric qos_latency is matching with AIQUM one.
Thanks, I will enable it and see what that is looking like. I’ll send you an update shortly
qos_latency metric seems to be exact same as AIQUM graph now. Really appreciate your help Rahul!
You are welcome. Differences between these 2 are documented in KB https://kb.netapp.com/on-prem/ontap/Perf/Perf-KBs/What_is_the_difference_in_latency_of_the_same_volume_between_statistics_volume_show_and_qos_statistics_volume_latency_show_commands