[gpfsug-discuss] gpfs performance monitoring

Alex Chekholko chekh at stanford.edu
Wed Sep 3 21:55:14 BST 2014

The usual way to do that is to re-architect your filesystem so that the 
system pool is metadata-only, and then you can just look at the storage 
layer and see total metadata throughput that way.  Otherwise your 
metadata ops are mixed in with your data ops.

Of course, both NSDs and clients also have metadata caches.

On 09/03/2014 10:27 AM, Salvatore Di Nardo wrote:
> Last but not least.. and this is what i really would like to accomplish,
> i would to be able to monitor the latency of metadata operations.
> In my environment there are users that litterally overhelm our storages
> with metadata request, so even if there is no massive throughput or huge
> waiters, any "ls" could take ages. I would like to be able to monitor
> metadata behaviour. There is a way to to do that from the NSD servers?

Alex Chekholko chekh at stanford.edu 347-401-4860

More information about the gpfsug-discuss mailing list