[Primex] Current Batch Farm jobs for primex account causing high nfs server loads

[ilarin at jlab.org]

 Dear David,
thanks for the attention for NFS traffic.
Please let me know if I understood you correctly:
You suggested not to use the standard batchjobs option INPUT_FILES,
since it jcaches,
but copy from mss to local farm disk like jget command doing?
Will this help to solve the nfs problem you a taking care of?

 Thanks again.
 Ilya

PS: I have noted that for a long time ifarm has a major problem this nfs
traffic, I'm pretty sure it has to be resolved and maybe by some
sofisticated way.


> Ilya,
> The current  primex batch farm jobs are putting extreme loads on our nfs
> cache fileservers. It appears that you are processing the files that
> have been cached for your farm job and staged onto the local disk of the
> farm node from the nfs copy of the file on the nfs cache fileserver.
> This puts an extreme load on the nfs cache servers as the farm nodes are
> processing the files from a network copy and not the file on the local
> disk of the farm node. Please modify your jobs so that the files being
> processed are not being processed from the network copy of the file (NFS
> mounted filesystem) but the copy of the file that is on the farm node
> disk as part of  your jobs.
>
>  Thank you for your attention.
>
> --
>
>
> -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> David Rackley         |      ********   **  **    **  ********  ********
> CC Sci Comp Sys Admin |        **      **  ***   **  **    **  **
> rackley at jlab.org      |       **      **  ** *  **  ********  ******
>                       |      **  *   **  **  * **  **    **  **
> Phone: 757.269.7041   |     **  ******  **   ***  **    **  **
> FAX:   757.269.6248   | TJNAF - Thomas Jefferson National Accelerator
> Facility
> -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>