John Hascall wrote:
>
> I was looking at the output of rxdebug -stats on our 5 main
> AFS servers and noticed a wide variation:
How did you compute these statistics?
> server\rtt: min avg max samples
> afs-1 0.000 0.434 24.271 15196228
> afs-2 0.000 0.175 16.272 28493566
> afs-3 0.000 0.236 48.968 26838336
> pfs-1 0.000 0.136 15.913 6621623
> pfs-2 0.000 0.052 16.232 7680353
>
> Are these numbers reasonable?
> Is the wide variation (esp the avg) something to worry about?
> What units are these in (I'm hoping not seconds!)?
Yes, they're in seconds. Note that the RTT is only computed during bulk
data transfer RPCs. I believe this is related to the other observations
about long-duration disk I/Os. "worry" may be a strong word, but it is
worth paying attention to.
