[j-nsp] NSR+GRES vs Graceful restart

[Mark Tinka]

On Thursday, February 24, 2011 09:06:37 pm Richard A 
Steenbergen wrote:

> But GRES+NSR is typically going to be a much better idea
> than GRES+GR.

I would tend to agree. 

Localization of faults and replication of state between 
control planes is, I think, a better idea than Graceful 
Restart.

> Of course I have yet to encounter an RE
> failure in the wild where GRES+anything actually saves
> the day,...

We've had situations where inserting a compact flash card 
into the router causes its CPU to spike so high, control 
planes protocols begin to complain, e.g., Graceful Restart 
has saved the day when LDP stops working for a while as the 
CPU "calms down", but MPLS forwarding is still going strong.

Of course, this is hardly a worst-case scenario, and in all 
fairness, most catastrophes we've had are generally 
controlled (upgrades, e.t.c.).

> but doing a hitless NSR switchover has been
> surprisingly helpful as a way to kick the box in
> response to a large number of random/obnoxious bugs,
> without actually having to disrupt forwarding.

Control plane switchovers, in our case, have benefited from 
GRES + Graceful Restart. BGP is the only thing that 
literally takes the longest to reconverge on new master RE, 
which is where we're thinking NSR will make the most sense.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20110226/165993ad/attachment.pgp>