We used to set an arbitrary qos limit which in some
cases happened to be higher than the actual disk
capabilities. Even though we had an explicit check
for that and we skipped the entire qos test suite
if the device was too slow, the disk performance could
vary and be just enough to pass that initial check,
but then slow down and fail in the middle of the test
suite. If the bdev maxes out at 21MB/s on one run, it
may just as well do 19MB/s on another. That is exactly
the case causing intermittent failures on our CI.
We fix it by removing the arbitrary qos limit and
setting it to a % of the maximum disk performance
instead. This lets us e.g. remove the code for skipping
the entire test suite when the disk is too slow. We
definitely don't want to skip any tests.
Change-Id: I6de8a183c00bab64484b4ddb12df1dedfbed23f8
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451887
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: yidong0635 <dongx.yi@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
22364ca8f7