Re: dracut requires router

[Seewer Philippe <philippe.seewer-omB+W0Dpw2o@public.gmane.org>]

On 10/08/2010 03:59 PM, Pádraig Brady wrote:
> On 08/10/10 14:30, Seewer Philippe wrote:
>> On 10/08/2010 12:20 PM, Pádraig Brady wrote:
>>> My terminals just failed to boot because the router
>>> was (temporarily) unavailable. Seems this is on purpose?
>>>
>>> http://dracut.git.sourceforge.net/git/gitweb.cgi?p=dracut/dracut;a=commit;h=98f25e96
>>>
>>>
>>> I don't know what the above is trying to do exactly.
>>
>> What this does ist simply try and give a possible router enough time to
>> update its STP caches (See
>> http://en.wikipedia.org/wiki/Spanning_tree_protocol) so that subsequent
>> mount attempts don't fail because the router isn't ready.
>>
>> So yes, it is on purpose if the available IP information contains routers.
>>
>> I Hope this helps
>
> Thanks for the info, but.
> Is it right to do this synchronously with such a long timeout?
It has to be done synchronously, sorry. We have to be sure the router is 
there and ready if the rootserver is behind the router.

As for the timeout: STP can take between 30-60 seconds. (There's a 
reason why there's newer protocols like RSTP etc).

> Is it right to die if we can't contact the router?
Usually yes. Either because the rootserver is behind the router or if 
you supplied routing information it's safe to assume that later you want 
it to work anyway.

But yes, usually. I would have preferred to somehow a subnet check on 
the rootserver to decide what or if to check. But alas, not all netroot 
variants know the ip of their rootserver at that point.

I'm always open for ideas on how to improve this.

Regards,
Philippe