Hi Peter,
Peter Tribble wrote:
> OK,
>
> A couple of things that confuse me about fsstat output:
>
> 1. Consider the following:
>
> % fsstat /
> new name name attr attr lookup rddir read read write write
> file remov chng get set ops ops ops bytes ops bytes
> 13.0K 1.44K 224 76.8M 1.79K 469M 426K 49.7M 12.9G 12.1M 2.53G /
>
> Now, for the numbers measured in bytes I understand that the
> multiplier is powers of 1024. I suspect (from comparing fsstat
> with raw kstat) that the same scaling is true of the plain
> operation counts. Does it make sense to use 1024 as the multiplier
> for these, where 1000 would be more natural.
>
> (I would expect factors of 1000, but then would get confused
> because different columns have different scaling factors...)
I gave some thought to this and decided that on balance, it was better
to have everything use the same scaling. (K=1024, M=1024K, etc.)
> 2. I'm running fssat and seeing the following for /
>
> new name name attr attr lookup rddir read read write write
> file remov chng get set ops ops ops bytes ops bytes
> 0 0 0 0 0 0 0 300 119K 215 121K /
> 0 0 0 31 0 119 0 292 119K 202 120K /
> 0 0 0 2 0 4 0 314 119K 246 121K /
>
> Huh? I checked with iostat and there's no I/O whatsoever.
>
> The culprit turns out to be a graphical java program sending
> it's X traffic back to my machine (tunneled over ssh). This
> goes via a socket, and reads and writes to that socket seem
> to get allocated to the / filesystem. I don't know where they're
> supposed to be accounted for (if at all), but it confused me
> for a while until I worked out what was going on.
>
Hmm... I'll look into that and see what's happening.
Thanks,
Rich
