Is it possible to set the performance data as the source data for harvest?
#Harvest monitor source data from Performance archive
drifting vortex
elfin cloud
@drifting vortex Harvest collects performance data from both ZapiPerf and RestPerf(9.12+) sources. Harvest does not support performance data collection from archive files. Are there any specific use cases for which you require the use of archive files?
drifting vortex
@elfin cloud Is there possibility to see the workloads data in the harvest?
I don't see the relative workload metrics from the default resource.
elfin cloud
@drifting vortex They are available but disabled by default
https://github.com/NetApp/harvest/blob/main/conf/zapiperf/default.yaml#L61-L64
They are disabled by default because they typically slow down data collection due to a high number of metrics.
drifting vortex
@elfin cloud The data I got from the metric is not the same to the data from PA "User Workload IOPS"
elfin cloud
@drifting vortex Could you add more details about the metric? When you say PA, do you mean AIQ PAS?
drifting vortex
@elfin cloud YES, AIQ PAS data
granite summit
thanks @drifting vortex can you share the exact metric name and we'll try to find it?
either by sharing a link to the dashboard or the panel json
elfin cloud
@drifting vortex Thanks. Harvest collects similar information and present it in Volume dashboard as in screenshot. Does this help?
drifting vortex
@elfin cloud Thanks for your help, but what I need is workload data, such as workload IOPS and throughput, instead of volumes
elfin cloud
@drifting vortex If you have enabled workload counters in Harvest and if you run query qos_ops in prometheus. Does it give what you are looking for?
similiarly for throughput qos_read_data and qos_write_data
drifting vortex
@elfin cloud Is the value get from metric "qos_read_ops" same as the value in PAS workload table view "read IOPS " ?
elfin cloud
@drifting vortex Harvest collects this metric from workload object in ONTAP. Description is as below
A workload represents work being done on behalf of an application or system process. The workload CM object reports information such as operations per second, a breakdown of where read operations are going, the interarrival time of operation request messages, working set size information, operation latency per workload, and deferred workload information. These statistics illustrate the filer's performance with given workloads
Counter description here
https://netapp.github.io/harvest/23.05/ontap-metrics/#qos_read_ops
While PAS workload dashboard seems to be focused only on workload_queue_nblade based on panel description.
The workload_queue_nblade object provides service center-based and delay center-based statistics about a workload from the nblade perspective. It includes a high level rollup of the CPU domain service and wait times from the nblade or protocol perspective, as well delay center statistics. These delay centers include cluster interconnect delay, qos limit delay, and network delay.
unique apex
It's slightly different objects but same calculation of data.
drifting vortex
@elfin cloud @unique apex Thanks for your explanation, I will discuss with the team based on user requirements on the monitor.
elfin cloud
@drifting vortex FYI. We are in process of making changes to workload dashboard as well https://github.com/NetApp/harvest/pull/2100