On Thu, Oct 21, 2010 at 03:57:21PM -0700, Mike Gerdts wrote:
> On Thu, Oct 21, 2010 at 10:18 AM, Ray Van Dolson <[email protected]> wrote:
> > Is there a way via kdb, kmdb or kstat to query the status of the
> > NFS/RPC duplicate request cache?
> >
> > I've tuned the size of mine up (cotsmaxdupreqs and maxdupreqs
> > variables) via mdb as a result of error messages, but would like to
> > keep an eye on how many entries are actually in this cache.
>
> How about dtrace.
>
> # dtrace -qn 'tick-1s { printf("%Y %d %d\n", walltimestamp,
> rpcmod`ndupreqs, rpcmod`cotsndupreqs); }'
>
> You can also access those variables with mdb. For example...
>
> # echo 'rpcmod`ndupreqs::print' | mdb -k
> 0
Thanks Mike.
The dtrace one looks like what I want. However, it's output is as
follows:
# dtrace -qn 'tick-1s { printf("%Y %d %d\n", walltimestamp,rpcmod`ndupreqs,
rpcmod`cotsndupreqs); }'
2010 Oct 26 08:50:11 0 8192
2010 Oct 26 08:50:12 0 8192
2010 Oct 26 08:50:13 0 8192
..
So if I'm reading this correctly, the duplicate request cache is
completely full? cotsmaxdupreqs is set to 8192 and cotsndupreqs is
reporting 8192 as well per above.
Not that being full is necessarily bad or unwanted, just making sure
I'm reading this correctly.
I'm thinking ndupreqs is just a counter incremented when a duplicate
request is received and not necessarily an indicator of how many
entries are being stored in the dup request cache currently...
Ray
_______________________________________________
nfs-discuss mailing list
[email protected]
