[Openais] Clarification on token loss timeout.

[Steven Dake]

On Thu, 2006-09-21 at 22:43 +0530, Shridhar Sahukar wrote:
> Hello Steve,
> 
> I have some clarification regarding the operations being done during
> token loss timeout.
> 
> When there is a token loss timeout, there could be 3 reasons:
> 1. Current interface to which the node is bound to might have failed.
> 2. There could be some issues in the network.
> 3. The other node has gone down.
> 
> In either cases, currently what is being done is, as soon as there is a
> token loss timeout, and if we are in OPERATIONAL state, then we invoke
> netif_check where we rebind all our sockets.
> 
> Other way could have been; we should first check if the current
> interface id DOWN. We should rebind all our sockets only if the
> interface has gone down. Otherwise I dont see any reason to do all the
> processing that is currently being done.
> 
> Could you please let me know if there are any other reasons for doing
> so?
> 

Thanks for copying the list.

It isn't necessary to rebind all of the interfaces.  It is done this way
now simply because that is how the code is implemented at the moment.
Part of the problem is down checking doesn't entirely work with redudant
ring which is a higher priority feature.

If you send me a patch to change it and it works, I'd be happy to merge
it.

Regards
-steve

> Regards,
> Shridhar
>