Open jcurryaz1 opened 5 years ago
We know the root cause of the issue. Also this happens only when tracing the RAID volume only on RHEL/CentOS kernels.
The issue is not trivial to fix, and for now we focus on other priorities.
@jcurryaz1 Could you retest this please and also share the fio config?
Traced information from VROC or drives used in a RAID volume are not reporting io.lba correctly and the latency is appearing to be zero
A simple RAID0 volume was constructed iotrace --start-trace --devices /dev/md126 -t 400 a fio mixed workload was started
No latency or qd was captured
Then repeat of workload but this time capture a trace of the 2 devices that were striped in the RAID volume iotrace --start-trace --devices /dev/nvme0n1,/dev/nvme1n1 -t 400 Now some latency is shown, iodepth is same as above
In both cases, the io.lba ranges from 0 to very large numbers
Please advise of a share to post the 115M and 220M trace files and large xls. I saved the 1st 20lines of the parsed output. parsed-md126-vroc-raid0-20lines.xlsx parsed-2drives-vroc-raid0-20lines.xlsx