Customer brought to my attention that at times, volume IOPs can exceed the node IOPs owning that volume, and if you sum all the volume iops of a partuclar node they can exceed the node IOPs value by a signifcant factor. Is there an explantion for this behavior?
1 - The volume IOPS are accounted for in the node that hosts the LIF for that volume. However in a cDOT cluster, the aggregate can be hosted in another node. As a result if the customer sums up the IOPS of the volumes with LIFs in a node rather than the volume’s aggregate in the node, then the IOPS should match with the reported numbers. In any case the total cluster IOPS needs to match with the sum of IOPS in all node clusters
2 – If the customer is doing a lot of metadata IOPS rather than READs and WRITEs, others have reported inconsistencies as well. This inconsistency lays with ONTAP as we just display counters made available to OCUM from ONTAP Counter Manager.
3 – The statistics of volumes and nodes are collected over a 5 minute period, so there may be time gaps between the volume measurement and node measurement. If that is even 1 second in a busy system then there is a discrepancy. In the long run, the counters match, but in a single measurement it may not.
Personally I'd point to 1. One tends to thing in terms of the node owning the data (volume) rather than the node serving the data (LIF). I fell into that trap, too.
If this post resolved your issue, please help others by selecting ACCEPT AS SOLUTION or adding a KUDO or both.