Sudden latency and CPU utilization from workloads going idle to busy resolved by QoS

Applies to

• ONTAP 9
• FAS/AFF systems

Issue

• CPU is high (near or at 100%) constantly
• WAFL_Ex may be one of the busiest domains (where Data Processing happens)
• Hosts report high IO wait times for a particular duration in a day.
• Clients report slowness while accessing multiple shares and could notice increased latency on AIQUM for same timestamps.

• Latency has increased and is higher suddenly
  • NFS may be impacted
• sysstat -M 1 can be used to check the busiest domain. 

Example: sysstat output shows CPU raising due to an increase in user workload (columns removed for readability)

Cluster::> node run -node <node> -command sysstat -x 1
 CPU    NFS   CIFS   HTTP   Total     Net   kB/s     HDD   kB/s
                                       in    out    read  write
 11%    1324     0      0    1324     169    131    5300      0
 28%      72     0      0      72     483    526    4928     12
 53%     175     0      0     175     254    407    5176     24
 23%     143     0      0     143     146     72    4752      0
 12%     230     0      0     230     134    259    5808     24
 40%    5766     0      0    5766     207    720   44336  36956
 53%     108     0      0     108   15698  14391   32340     24
 46%      30     0      0      30   30975  30269   29900      0
 87%   32124     0      0   32124  576397  53287  203513     12
 99%   44334     0      0   44334  659406  45518  256931 251353
 99%   43692     0      0   43692  609739  16930  263599 565448
 99%   44492     0      0   44492  633509  41562  261366 116257