Peak performance latency over 4 at every Saturday with no workload | NetApp | Page 1

lilac bear Jul 26, 2023, 4:14 AM

#

Hi, I have a node A250 that has a peak latency that happens every Saturday at midnight, but the node has no workload, and the efficiency rate is 1.
What could be the reason for that peak latency?
The normal latency is around 1 and the peak point is around 4

gaunt bridge Jul 26, 2023, 7:58 AM

#

hard to say with the little info you gave. I can suggest two things. a) install NAbox to monitor the system and dig into the charts and graphs it provides, or b) opena NetApp case and be prepared to collect performance ASUPs

#

but if there is "no workload" at this time, does the latency increase actually matter? i.e. if nobody is driving on the road, does it matter if it has only one or two lanes instead of 6? 😉

lilac bear Jul 26, 2023, 2:56 PM

#

gaunt bridge hard to say with the little info you gave. I can suggest two things. a) install ...

@gaunt bridge I understand that the peak performance can be ignored if there is no workload, but the user might still be confused about that, is it possible to be system doing some job itself?

gaunt bridge Jul 26, 2023, 5:37 PM

#

what I have seen a few times in the past is that, if there is very low activity on the system, latencies are much higher than normal and go down if you put more load on the system. Might be a statistics thing or something, I don't know

tepid rose Jul 26, 2023, 8:02 PM

#

There are many background jobs in ONTAP with the current versions, like: filesystem analytics, background efficiency scans (deduplication, compression, compaction), ARP (autonomous ransomware protection), WAFL scans like IDR (inactive-data-reporting), or FabricPool cooling, disk scrubs, etc.
But all of these are generally throttled if user workloads need the resources (performance headroom is being checked constantly).

Off the top of my head I don't know of anything which happens regularly every Saturday. Maybe check if some snapmirrors or efficiency scans are scheduled at that time. Is Fpolicy and/or VScan configured? Could also lead to high latencies.

With "the node has no workload" you mean there are no clients accessing the system via data protocols (SMB, NFS, iSCSI, etc.) at that time? Or does it mean there are basically no volumes on the system?

This might help: https://kb.netapp.com/onprem/ontap/Performance/How_to_check_for_background_CPU_utilization_in_ONTAP_9