[K12OSN] nfs troubles

[Les Mikesell]

Roger wrote:
> On Nov 16, 2007 12:42 PM, Les Mikesell <les at futuresource.com> wrote:
>> Roger wrote:
>>> On Nov 16, 2007 12:08 PM, "Terrell Prudé Jr." <microman at cmosnetworks.com> wrote:
>>>>  Yep, I can see that happening.  Remember that, with IPX/SPX, the network
>>>> address is done by MAC address (not so with IPv4), and the addresses are
>>>> thus auto-assigned (no DHCP).  IPv6 "borrowed" this little trick and thus
>>>> doesn't need DHCP any more than IPX does.
>>> yep remembered the addressing. this was 5 or 6 years ago, so I wasn't
>>> really sure what the problem was.  it did have something to do with no
>>> IP address.   I think what the user saw was getting a login to the
>>> file server, but their mail wouldn't work or something.  that's when
>>> we got to digging around and found out that pesky delay.   I think
>>> spanning tree has two or three 30 second delays during the port
>>> enabling process.   A listening phase, a learning phase, something
>>> else, then it enables the port.
>>> portfast essentially bypasses the listening and learning.  (dang, is
>>> that the name?)..  It just enables the port.    It does enable
>>> spanning tree, just the pre-enabling checks aren't done.
>> Portfast doesn't really stop the spanning tree process - it just
>> forwards packets during this step.  It will still shut the port down if
>> it learns it is a loop (unless there is so much traffic that it drops
>> the packets it needs to see that...).  Just remember never to plug
>> another switch into a port you have set to portfast mode and you won't
>> have loops.
> 
> You can still have switches plugged in, just the initial check isn't
> done.  If you have a loop when you plug it in, it can be a problem.
> If you have a switch plugged in and create a loop later on, spanning
> tree will catch it (supposed to).
> right?  or did I get that part wrong?

The point of spanning tree is that you are supposed to be able to have 
loops (otherwise known as redundant connections...) and only one path 
should forward at a time.  The whole computation has to be done every 
time an interface goes up or down or a mac address is seen from a new 
direction.  The portfast directive doesn't stop the computation, it just 
  allows packets to go through while it happens instead of blocking 
until it is sure it is supposed to forward.

-- 
   Les Mikesell
    les at futuresource.com