[Open-FCoE] [PATCH] libfc: rport retry on LS_RJT from certain ELS
jre at nuovasystems.com
Fri Jan 16 00:14:17 UTC 2009
Chris Leech wrote:
> Joe Eykholt wrote:
>> I have seen this situation here, even with 2 second timeouts. We'd get
>> multiple rejects so maybe we're not retrying enough times.
> Maybe we should try an exponential back-off in the case of multiple
> rejects? That way we can start with a very short delay, but not hammer
> the target with requests if it's actually going to take a long time.
Maybe eventually, if it seems necessary.
I can see it starting at E_D_TOV for the first few retries and
then multiply it by 1.5 or 2 every retry up to a max of say 32 * E_D_TOV?
But this would also be limited by the retry count limit.
I think the 2 second default E_D_TOV isn't going to cause too much
traffic as it is, so increasing the number of retries ought to be enough.
More information about the devel