【Problem Description】: After changing the active copy of the Database to s00lion01nod (previously it wass00lion02nod) 17.06.2020-we began to observe a periodic, short-term increase in the response of the disk system, such a response is not acceptable for this one.
【Problem Analysis】 : About the large latency during 11:10-11:30
We found that all luns which were mapped to s00lion01nod host were in the same lun group which ID is 1 .
Also there is a smartqos policy configured for lun group 1 as bellow :
Max IOPS configured 200000 and max BandWidth configured 3000
Then we checked if smartqos limited the band width.
Also we can see here, the IOPSdid not reach the 200000, but we need to notice that when we configure IOPSlimitation in a smartqos policy, we can see below :
Different IO size, different Max IOPS.200000 is a value for 8KB IO. But in fact, during that time period, the size of many IOs was larger than 512KB.
Therefore, the IOPS limit is adjusted in realtime as the I/O size changes.
Conclusion: IOPS was limited by smartqos on storage during that time, then the latency increased on the host side.
【Root Cause】: IOPSwas limited by smartqos on storage during that time, then the latency increased on the host side.
【Solution Description】: two ways to resolve this issue.
Increase the IOPS limitation of smartqos policy for this lun group.
Cancel IOPS limitation, only limit the bandwidth.
