[j-nsp] cgnat on service module - interesting bgp advertisements

Aaron aaron1 at gvtc.com
Wed Apr 20 13:59:49 EDT 2016


Ok thanks… yep I’m seeing them getting used….

 

agould at eng-lab-mx104-cgn> show services sessions | grep 1.2.3.128

TCP        52.26.119.1:80    ->   1.2.3.128:1053  Forward  O               4

TCP       64.12.245.38:80    ->   1.2.3.128:1052  Forward  O               5

ICMP           8.8.8.8       ->   1.2.3.128       Forward  O            6873

 

agould at eng-lab-mx104-cgn> show services sessions | grep 1.2.3.255

TCP     54.240.170.135:80    ->   1.2.3.255:1052  Forward  O               7

TCP     2.4.178.10:80        ->   1.2.3.255:1051  Forward  O               9

 

 

Aaron

 

 

From: Alexander Arseniev [mailto:arseniev at btinternet.com] 
Sent: Wednesday, April 20, 2016 11:58 AM
To: Aaron <aaron1 at gvtc.com>; juniper-nsp at puck.nether.net
Subject: Re: [j-nsp] cgnat on service module - interesting bgp advertisements

 

Hello,
When using "address X.Y.Z.W/TV" definition, the net and bcast IPs are NOT used and JUNOS carves them out.
When using "address-range low X.Y.Z.W high X.Y.W.U", ALL addresses are used and JUNOS does NOT carve the net and bcast IPs out.
Whether You _should_ use net and bcast IPs - it depends on the app mix. 
Some legacy apps such as eMule frown upon IPs ending with 0 (i.e. net IP for /24), and gives them "low id"
https://en.wikipedia.org/wiki/EMule#Low_ID 
HTH
Thanks
Alex

On 20/04/2016 17:11, Aaron wrote:

Awesome Alex, good find !

 

So I wonder if the high-cpu issue with the all-zero and all-ones address of the named subnet means that when using address-range that the net-id and bcast address for the low/high definition will NOT be used in the nat pool ?  …or that net-id and bcast addresses WILL be used for nat translations ?

 

Thanks again

 

 

agould at eng-lab-mx104-cgn# show | compare

[edit services nat pool nat1]

-     address 1.2.3.128/25;

[edit services nat pool nat1]

+     address-range low 1.2.3.128 high 1.2.3.255;

 

 

 

agould at eng-lab-mx104-cgn# run show route advertising-protocol bgp 10.101.0.2 table one.inet.0

 

one.inet.0: 771 destinations, 1915 routes (771 active, 0 holddown, 0 hidden)

  Prefix                  Nexthop              MED     Lclpref    AS path

* 10.144.2.4/30           Self                         100        I

* 1.2.3.128/25         Self                         100        I

 

[edit]

 

 

 

 

From: Alexander Arseniev [mailto:arseniev at btinternet.com] 
Sent: Wednesday, April 20, 2016 1:36 AM
To: Aaron  <mailto:aaron1 at gvtc.com> <aaron1 at gvtc.com>; juniper-nsp at puck.nether.net <mailto:juniper-nsp at puck.nether.net> 
Subject: Re: [j-nsp] cgnat on service module - interesting bgp advertisements

 

Hello,
MS-MIC (and MS-MPC NPUs as well) automatically cuts out network (in your case .128) and broadcast (in your case .255) IPs.
The rest cannot be expressed as single prefix, hence a bunch of smaller prefixes is annonced instead.
This was done as PR 1019354 fix
https://prsearch.juniper.net/InfoCenter/index?page=prcontent <https://prsearch.juniper.net/InfoCenter/index?page=prcontent&id=PR1019354> &id=PR1019354 
HTH
Thx
Alex

On 20/04/2016 00:48, Aaron wrote:

Very interesting. anyone know why this is happening ?  Is this documented ?
I put a /25 as the public nat pool, but look what this mx104 is advertising
via bgp.. It appears to chop up that /25 into a bunch of smaller subnets and
advertise those out
 
 
 
 
 
agould at eng-lab-mx104-cgn> show configuration | grep 1.2.3. | display set
 
set services nat pool nat1 address 1.2.3.128/25
 
 
 
 
 
agould at eng-lab-mx104-cgn> show route advertising-protocol bgp 10.101.0.2
table one.inet.0
 
 
 
one.inet.0: 782 destinations, 970 routes (782 active, 0 holddown, 0 hidden)
 
  Prefix                  Nexthop              MED     Lclpref    AS path
 
* 10.144.2.4/30           Self                         100        I
 
* 1.2.3.129/32         Self                         100        I
 
* 1.2.3.130/31         Self                         100        I
 
* 1.2.3.132/30         Self                         100        I
 
* 1.2.3.136/29         Self                         100        I
 
* 1.2.3.144/28         Self                         100        I
 
* 1.2.3.160/27         Self                         100        I
 
* 1.2.3.192/27         Self                         100        I
 
* 1.2.3.224/28         Self                         100        I
 
* 1.2.3.240/29         Self                         100        I
 
* 1.2.3.248/30         Self                         100        I
 
* 1.2.3.252/31         Self                         100        I
 
* 1.2.3.254/32         Self                         100        I
 
 
 
 
 
Aaron
 
 
 
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net <mailto:juniper-nsp at puck.nether.net> 
https://puck.nether.net/mailman/listinfo/juniper-nsp

 

 



More information about the juniper-nsp mailing list