Re: [j-nsp] maxium number of rvi's on ex series?

2010-02-15 Thread Dan Farrell
In full disclosure we hit a limit on 10.0.R2.10 with the number of AE links we 
could provision- 10. Additionally we saw the memory link error that is 
(apparently now) characteristic of running that code- we are now at 10.0.S1.1 
(as advised by JTAC) and have 12 AE links working just fine. At this point we 
use interface-ranges so heavily that it would be a PITA to take the time to 
reconfigure everything without the ranges (and then reconfigure with them when 
we decide to go back to the 10.x train).

I'm not sure where (version) the memory leak is introduced in JUNOS but it's 
between the marvel chipset in the PFE and the RE. I think it was also limiting 
some of our VLAN creation.


Dan
-Original Message-
From: Ross Vandegrift [mailto:r...@kallisti.us]
Sent: Friday, February 12, 2010 8:45 PM
To: Dan Farrell
Cc: juniper-nsp@puck.nether.net
Subject: Re: [j-nsp] maxium number of rvi's on ex series?

On Fri, Feb 12, 2010 at 07:51:53AM -0800, Dan Farrell wrote:
> Are there any hard limits that anyone knows of? We use the 3200's and
> 4200's, and on the 4200's we're literally putting on hundreds of rvi's
> (eventually a couple thousand).

I've been told that the only limitation is the FIB size, or every VLAN (since 
the EX doesn't support multiple bridge domains).

In practice, if you're doing that, it implies you might have a large
L2 config.  We have hit a number of bugs related to the CPU time it takes to 
commit a large L2 config.  Turns out that can spin enough to interfere with 
periodic packet processing.

So - it should absolutely work.  You definitely want to be running
9.6R3 as it has a fix for a potentially serious scheduling bugs related to 
large layer 2 configs.  It's been decided that this fix will NOT be backported 
to 9.3.

Ross

--
Ross Vandegrift
r...@kallisti.us

"If the fight gets hot, the songs get hotter.  If the going gets tough, the 
songs get tougher."
--Woody Guthrie


__ Information from ESET NOD32 Antivirus, version of virus signature 
database 4869 (20100215) __

The message was checked by ESET NOD32 Antivirus.

http://www.eset.com

___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp


Re: [j-nsp] maxium number of rvi's on ex series?

2010-02-12 Thread Ross Vandegrift
On Fri, Feb 12, 2010 at 07:51:53AM -0800, Dan Farrell wrote:
> Are there any hard limits that anyone knows of? We use the 3200's
> and 4200's, and on the 4200's we're literally putting on hundreds of
> rvi's (eventually a couple thousand).

I've been told that the only limitation is the FIB size, or every VLAN
(since the EX doesn't support multiple bridge domains).

In practice, if you're doing that, it implies you might have a large
L2 config.  We have hit a number of bugs related to the CPU time it
takes to commit a large L2 config.  Turns out that can spin enough to
interfere with periodic packet processing.

So - it should absolutely work.  You definitely want to be running
9.6R3 as it has a fix for a potentially serious scheduling bugs related
to large layer 2 configs.  It's been decided that this fix will NOT be
backported to 9.3.

Ross

-- 
Ross Vandegrift
r...@kallisti.us

"If the fight gets hot, the songs get hotter.  If the going gets tough,
the songs get tougher."
--Woody Guthrie


signature.asc
Description: Digital signature
___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

[j-nsp] maxium number of rvi's on ex series?

2010-02-12 Thread Dan Farrell
Are there any hard limits that anyone knows of? We use the 3200's and 4200's, 
and on the 4200's we're literally putting on hundreds of rvi's (eventually a 
couple thousand).



Thanks,



Dan Farrell

Director of Network Operations

Applied Innovations Corp.

da...@appliedi.net



___
juniper-nsp mailing list juniper-nsp@puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp