NetApp-Harvest(Per second latency stats)

[Debbrata]

We have a requirement from a Banking customer in APAC to have the volume latency & ops stats per second.

They want to get the data for few critical volumes only(Not all volumes in the system)

The data retention is required for 24 hours , post which it can be rolled over to the next day.

The application is quite latency sensitive and they would experience latency which lasts for few seconds and hence they would like to have a mechanism to view the granular level latency from Storage end.

We would like to understand the feasibility of getting this configured and  the space requirement for the same.

Also would like to understand , if this would have any repercussion on the system due to excessive data polling (like high CPU utilization due to Host_OS etc.)

[paul_stejskal]

https://kb.netapp.com/mgmt/AIQUM/How_to_increase_performance_data_samples_in_Active_IQ_Unified_Manager may be related?

[Debbrata]

Can this be executed per instance , as the requirement is to have granular stats for few critical volumes only.

[paul_stejskal]

Deb,

Not sure. Maybe check with the Harvest folks on Discord? I work quite a bit with them and they are very helpful.

I don't think AIQUM can do it in current form, but Harvest might be able to (may require an XML to change perf archive counter frequency?).

[paul_stejskal]

Now that I think about it, it may be easier to run a script that checks for one volume only if that is the requirement.

[Ontapforrum]

Have you tried AIUM Performance stats ?

How to monitor volume latency from Active IQ Unified Manager
https://kb.netapp.com/mgmt/AIQUM/How_to_monitor_volume_latency_from_Active_IQ_Unified_Manager

1) As Kb suggests - Under Volumes pane, in the View drop down, click "Performance - All Volumes".
2) Select any one volume (i.e one you need)
3) It will take you to 'Explorer' tab in the top menu
4) In the VIEW & COMPARE, you can select either 'Volumes on same Aggregate' or select another option.
5) On the Right-hand side pane, there is TIME RANGE: select 'LAST 24 Hours' and click 'Apply Range'.
6) Now, on the LEFT pane - You can view avg IOPS/Latency/MB/sec for last 24 hrs, and on the RIGHT PANE, you can actually toggle along the time-line see any sharp PEAKS/lows and view the stats.

Method mentioned above, is quite handy (to assess the acceptable performance per se), but if you need more granular data-perf-stats  then only thing that I can think of is to run the QoS cmd directly into the console via putty (change settings | logging | save it on a drive which is large enough, b'cos that notepad will grow quite steadily, from 4kb to 8kb to 16kb to 32kb to 64kb etc)

https://kb.netapp.com/onprem/ontap/Performance/How_to_use_qos_statistics_commands_monitor_volume_latency_real_time

::> qos statistics volume latency show -node <node> [Across all components cluster/data/network/disk etc]
::> qos statistics volume performance show -vserver <vserver> -volume <volume> [Just IOPS/Latency/Throughput]

If you want to keep an eye on the CPU then, use this:
::> qos statistics resource cpu show

[Debbrata]

Hi Ashwin ,

Even though the stats mentioned above are per second , the sample is for 5 minutes. That does not serve customer's requirement of monitoring the latency per second as latency burts for few seconds could get averaged out over the 5 minute sample.

They would want to view these stats via Harvest using per second polling.