[j-nsp] VRRP tuning and scaling

[Gard Undheim]

Hi

What Junos version are you running, I have currently 100 vrrp
groups, all with fast-interval 100msec. This config committed
without any warnings. All vrrp-groups are under the
same interface and unit. This is on a M5 running Junos 8.5.

Config:

fe-0/2/2 {
     description to-3550;
     vlan-tagging;
     unit 99 {
         vlan-id 99;
         family inet {
             address 10.10.10.3/24 {
                 vrrp-group 1 {
                     virtual-address 10.10.10.1;
                     priority 90;
                     fast-interval 100;
                     accept-data;
                 }
                 vrrp-group 10 {
                     virtual-address 10.10.10.10;
                     priority 90;
                     fast-interval 100;
                     accept-data;
                 }
                 vrrp-group 11 {
                     virtual-address 10.10.10.11;
                     priority 90;
                     fast-interval 100;
                     accept-data;
                 }
                 vrrp-group .. {
                     virtual-address 10.10.10...;
                     priority 90;
                     fast-interval 100;
                     accept-data;
                 }
		vrrp-group 99 {
                     virtual-address 10.10.10.99;
                     priority 90;
                     fast-interval 100;
                     accept-data;
                 }

Regards,
Gard

Angel Bardarov wrote:
> Hi,
> When you configure "fast interval" you have limitation on number of VRRP 
> groups - it depends  also on how "fast" is your interval:
> 
> vrrp-group 250 {
>     virtual-address 1.1.1.252;
>     priority 105;
>     fast-interval 100;
>     preempt {
>         hold-time 300;
>     }
>     accept-data;
> }
> [edit interfaces ge-0/2/1 unit 2011 family inet address 1.1.1.3/24]
> lab at edge-a1-RE0# commit check
> [edit interfaces ge-0/2/1 unit 2011 family inet address 1.1.1.3/24]
>   'vrrp-group 20'
>     Configuration exceeds maximum rate of 200 fast advertisements per second
> error: configuration check-out failed
> 
> With 100 ms interval the convergence is around 400 ms and you can't have 
> more than 20 groups. You can have more than 20 if these additional 
> groups don't use "fast interval" feature.
> 
> Regards,
> Angel
> 
> 
> Gard Undheim wrote:
>> Hi
>>
>> Thanks for the answers. I have tested a bit with up to 75 vrrp-groups
>> and a interval of 100msec in the lab now, and the boxes is not able to
>> to the failover in the expected time (300-400msec). These tests were
>> done with two M5's, the actual network will use M120 and M320. I guess
>> I'll set the interval to 300 seconds, since this is also what is
>> recommended in the 8.5 release notes.
>>
>> Regards,
>> Gard
>>
>> Jonathan Brashear wrote:
>>   
>>> Seconded.  Depending on what type of Juniper you're using, VRRP issues
>>> can heavily tax a router and cause it to drop bgp with other routers if
>>> enough problems pop up.  Dropping the fast-interval time will only
>>> hasten these problems.  It's your call, but imo it's a
>>> high-risk/low-reward situation tuning the failover to the lowest time
>>> possible. 
>>>
>>>
>>> Network Engineer
>>>     
>>>> 214-981-1954 (office) 
>>>> 214-642-4075 (cell)
>>>> jbrashear at hq.speakeasy.net 
>>>>       
>>> http://www.speakeasy.net
>>> -----Original Message-----
>>> From: juniper-nsp-bounces at puck.nether.net
>>> [mailto:juniper-nsp-bounces at puck.nether.net] On Behalf Of alaerte vidali
>>> Sent: Sunday, January 13, 2008 4:04 PM
>>> To: juniper-nsp at puck.nether.net
>>> Subject: Re: [j-nsp] VRRP tuning and scaling
>>>
>>> Hi,
>>>
>>> I believe you will face scalability issues, even though I can only
>>> affirm
>>> this 100% in Cisco.
>>>
>>> I suggest you consider not only normal periods of the network. In this
>>> case
>>> probably it is ok.
>>> But add to this all process you have, plus things like instabilities in
>>> network (STP issues, link flapping...).
>>> Then everything together can result in routers losing track of VRRP
>>> neighbor
>>> packets due to the large number.
>>>
>>> Br,
>>> Alaerte
>>>
>>>     
>>>> Today's Topics:
>>>>
>>>>   1. VRRP tuning and scaling (Gard Undheim)
>>>>
>>>>
>>>> ----------------------------------------------------------------------
>>>>
>>>> Message: 1
>>>> Date: Sat, 12 Jan 2008 18:15:58 +0100
>>>> From: Gard Undheim <gard at ipnett.no>
>>>> Subject: [j-nsp] VRRP tuning and scaling
>>>> To: juniper-nsp at puck.nether.net
>>>> Message-ID: <4788F5CE.5090905 at ipnett.no>
>>>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>>>
>>>>
>>>> Hi
>>>>
>>>> I am currently tuning VRRP to speed up the failover time. I have set
>>>>       
>>> the
>>>     
>>>> fast-interval to 100msec, which is a minimum. Does anyone know how
>>>>       
>>> these
>>>     
>>>> settings scale with different numbers of vrrp groups? The customer
>>>> currently has 20 vrrp groups, but would it also work fine with as much
>>>> as 255 groups and fast interval of 100msec?
>>>>
>>>>
>>>> Thanks for any help!
>>>>
>>>> Best Regards,
>>>> Gard
>>>>
>>>>
>>>>
>>>>       
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>> _______________________________________________
>>> juniper-nsp mailing list juniper-nsp at puck.nether.net
>>> https://puck.nether.net/mailman/listinfo/juniper-nsp
>>>     
>>
>>   

-- 

Best Regards,
Gard Undheim
-----------------------------------------
Systems Engineer, Network Solutions
IPnett AS
-----------------------------------------
Phone:  +47 67 20 10 10 (Direct -63)
Fax:    +47 67 20 10 11
Mobile: +47 97 72 86 71
Email:  gard at ipnett.no
=========================================