Q:We are wondering whether the "latency" column shown by the command "zpool iostat <pool> 1 100" is reliable enough to be taken into account for troubleshooting. For instance, in the following paste shows max values between 700 - 750 (ms?) which we think
are very high.
https://hig.eis.utoronto.ca/pwd/?938aeeab597fb5ed#B5j0hrcT/4+9DZ6lXLk9wnZNy2WwItSD0Jox7GMDLVc=
Would you please clarify? I have search documentation about this topic online and couldn't be able to find anything. This column seems to be part of ZFS only in Nexenta, since I couldn't see it neither in Solaris 11 nor OpenIndiana 11.
A:Yes, the numbers you see there are meaningful. I will caution that they're isolated to a specific layer, so they don't fully explain what is going on and why. You may find that `zpool iostat -v` gives you more informative information (per-disk) with the
same caveat applied. `iostat -xn` may also be informative, although here again I will caution that the %b number which can often hit "100%" can easily be misinterpreted -- the service time columns are generally more representative and easier to interpret.
The `zpool iostat` functionality is present in most or all modern OpenZFS implementations.