We had a huge write latency( approx 200 msecs) in all volumes of a specific SAS aggr, let's say aggr1 .Its having a 47 disks with 3 RG -16,16,15 .
Aggr IOPS & Volume IOPS are very low been around 150 .
CPU Utilization was normal being 50% ---Is there any way to calculate the NVRAM usage ? as its a write latency i suspect perhaps it could be any reason.
Can anyone guide me, what are the areas i need to check /look to mitigate this ?
More over i don't believe it could be a loop saturation since its a SAS loop ,anyone correct me if i wrong.
Solved! See The Solution
Ideally gather a perfstat and send to NetApp support, but for a quick check in the meantime...
During the period of high latency, run a sysstat -x 1 on the console. Firstly look at the 'CP ty' column, watch out for any 'B' or (worse) 'b'. This suggests your NVRAM is not big enough, or you're not flushing data to disk fast enough (busy disks). Also for a rough disk utilisation figure, check the 'Disk Util' column to see how busy the busiest disk on the system is.
If the problem is NVRAM being full ('B' or 'b') you would normally expect ALL write workloads on the filer to be affected, not just those on the affected aggr.
To see how busy the disks are, you can use statit.
Hope that helps,