On Wed, Sep 30, 2009 at 5:07 PM, Michael Meffie <[email protected]> wrote:
> Hello,
>
> I would like to propose a change to the rx_statistics
> structure which is returned in RX debug packets. The proposed
> change would append two integers to the rx_statistics to
> report the number of ack packets that have been delayed due to
> excessive aborts (rx throttling).  (See the fileserver
> --abortthreshold option and related rx code.)
>
>
>  struct rx_statistics {         /* General rx statistics */
>     int packetRequests;                /* Number of packet allocation
> requests */
>     int receivePktAllocFailures;
>     int sendPktAllocFailures;
>     int specialPktAllocFailures;
> ...etc...
>     int receiveCbufPktAllocFailures;
>     int sendCbufPktAllocFailures;
>     int nBusies;
> -    int spares[4];
> +    int delayedConnAbortsSent;  /* Number of delayed connection aborts sent
> (throttling) */
> +    int delayedCallAbortsSent;  /* Number of delayed call aborts sent
> (throttling) */
> +    int spares[2];
>  };
>
> The number of spares is reduced to 2 integers by this change,
> but since this is not an RX call, but rather the payload of an
> rx debug packet, I believe addtional integers could be added
> up to the limit imposed on the packet size.
>
> In terms of the standardization process, it seems this would
> best live in an updated rx-specification. The draft document
> by Nickolai Zeldovich only refers to the file rx/rx.h for the
> contents of the rx debug packet.  See the section "Debugging"
> in http://web.mit.edu/kolya/afs/rx/rx-spec.

I'm not sure that's the best practice for doing so, though for this an
I-D is probably too heavyweight. Regardless of how we decide we'd like
to document this, I am in favor of this proposal.

-- 
Derrick

_______________________________________________
AFS3-standardization mailing list
[email protected]
http://michigan-openafs-lists.central.org/mailman/listinfo/afs3-standardization

Reply via email to