[j-nsp] Loopback IP address in BGP Peering

[chenoi A]

Hai...just for understanding
loopback is configure in eBGP setup. Loopback is virtual interface..
a. in configuring bgp do we need to have loopback
b. does it using private or public ip
c. in a block of IP /22 how many ip that we need to reserve for loopback
d. if i configure one interface with 211.1.221.3.1/30 , does the loopback using the same ip or others ip.
TQ in advance.

[Mark Tinka]

Post by chenoi A
Hai...just for understanding
loopback is configure in eBGP setup.

A Loopback interface is configured for several reasons,
including (but not limited to) BGP.

Loopback interfaces will typically be used on eBGP setups
for load balancing over multiple links to an upstream, or
for eBGP multi-hop.

I tend to discourage eBGP multi-hop for a number of reasons,
the most being it adds much unneeded complexity and
potential for problems you'd much rather avoid.

Post by chenoi A
Loopback is virtual
interface.. a. in configuring bgp do we need to have
loopback

A Loopback is great for:

* iBGP sessions
* Router ID.

Post by chenoi A
b. does it using private or public ip

That doesn't matter. As long the address is routed within
your network.

However, for commercial Internet services, recommended all
your interfaces be numbered out of public address space,
including your Loopback interface.

Post by chenoi A
c. in a block of IP /22 how many ip that we need to
reserve for loopback...

Depends on how many routers you plan to have throughout your
network.

Remember, your Loopback address takes a netmask of /32. So
you can have as many address as you (plan to) have
routers... with the exception of x.x.x.0 and x.x.x.255, of
course.

Given the size of your allocation, I'd start off with a /25
for all Loopbacks and slice it further based on how many
routers you plan to have in each PoP. But this is your call,
don't take my word for it :-).

Post by chenoi A
d. if i configure one interface with
211.1.221.3.1/30 , does the loopback using the same ip or
others ip. TQ in advance.

No, the Loopback interface generally has its own IP address.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20090228/1dd70098/attachment.bin>

[Cougar]

Post by Mark Tinka
Remember, your Loopback address takes a netmask of /32. So
you can have as many address as you (plan to) have
routers... with the exception of x.x.x.0 and x.x.x.255, of
course.

What kind of exception is this? In CIDR world you can use any address you
like except first and last _LAN_ addresses when netmask is /30 or less.
With /31 and /32 can use any address and so far I haven't seen any
problems using x.x.x.0 or x.x.x.255 in Junipers.

---
Cougar

[Mark Tinka]

Post by Cougar
What kind of exception is this? In CIDR world you can use
any address you like except first and last _LAN_
addresses when netmask is /30 or less. With /31 and /32
can use any address and so far I haven't seen any
problems using x.x.x.0 or x.x.x.255 in Junipers.

That may very well be - but my suggestion is just because it
can be done, doesn't mean it's a great idea "all around".
These are the types of practices that come back and bite you
due to varying levels of support for implementing .0 and
.255 across various pieces of software. I'm not presuming
the OP has only Junipers to deal with in their network.

Given the number of addresses one may potentially save in,
say, a /24 sliced only for Loopbacks vs. not getting
stressed by why this may break some things in the network;
I'd much rather sacrifice those two addresses, thank-you-
very-much.

Keep it simple, keep it stupid, keep it unambiguous. The
physics don't change, just how you apply them.

Then again, to each his own...

You probably want to spend some time wading through:

http://tinyurl.com/dzw4cj
http://tinyurl.com/av8rwm
http://tinyurl.com/chwjms

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: This is a digitally signed message part.
URL: <https://puck.nether.net/pipermail/juniper-nsp/attachments/20090228/18f5212a/attachment.bin>

[Patrik Olsson]

All major vendors support 0 and 255 as loopback in the last octet. o if
you are using J or C for instance, your safe. If you use an obscure
small vendor, the choice to use 0 and 255 can come back and bite you in
the ***.

Cheers
Patrik

Post by Mark Tinka

Post by Cougar
What kind of exception is this? In CIDR world you can use
any address you like except first and last _LAN_
addresses when netmask is /30 or less. With /31 and /32
can use any address and so far I haven't seen any
problems using x.x.x.0 or x.x.x.255 in Junipers.

That may very well be - but my suggestion is just because it
can be done, doesn't mean it's a great idea "all around".
These are the types of practices that come back and bite you
due to varying levels of support for implementing .0 and
.255 across various pieces of software. I'm not presuming
the OP has only Junipers to deal with in their network.
Given the number of addresses one may potentially save in,
say, a /24 sliced only for Loopbacks vs. not getting
stressed by why this may break some things in the network;
I'd much rather sacrifice those two addresses, thank-you-
very-much.
Keep it simple, keep it stupid, keep it unambiguous. The
physics don't change, just how you apply them.
Then again, to each his own...
http://tinyurl.com/dzw4cj
http://tinyurl.com/av8rwm
http://tinyurl.com/chwjms
Mark.
------------------------------------------------------------------------
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

[C T]

Just to add a comment.
Some people confuse it, including?me sometimes... :-)
?
Even a router or a system with Classfull (without CIDR) the address z.y.x.0 or z.y.x.255 can be considered valid host addresses.
?
For instance, if you have a class B 128.3.0.0 (CIDR equivalent of 128.3.0.0/16)
The addresses:
128.3.10.0 is a valid host address
128.3.10.255 is a valid host address
?
But using class C, the .255 and .0 are invalid host addresses.
?
Good reference at: http://tools.ietf.org/html/rfc4632
?
Cheers,
Claus T
?

All major vendors support 0 and 255 as loopback in the last octet. o if >you are using J or C for instance, your safe. If you use an obscure >small vendor, the choice to use 0 and 255 can come back and bite >you in the ***.
Cheers
Patrik

Post by Cougar
What kind of exception is this? In CIDR world you can use any address
you like except first and last _LAN_ addresses when netmask is /30 or
less. With /31 and /32 can use any address and so far I haven't seen
any problems using x.x.x.0 or x.x.x.255 in Junipers.

That may very well be - but my suggestion is just because it can be
done, doesn't mean it's a great idea "all around".
These are the types of practices that come back and bite you due to
varying levels of support for implementing .0 and
.255 across various pieces of software. I'm not presuming the OP has
only Junipers to deal with in their network.
Given the number of addresses one may potentially save in, say, a /24
sliced only for Loopbacks vs. not getting stressed by why this may
break some things in the network; I'd much rather sacrifice those two
addresses, thank-you- very-much.
Keep it simple, keep it stupid, keep it unambiguous. The physics don't
change, just how you apply them.
Then again, to each his own...
http://tinyurl.com/dzw4cj
http://tinyurl.com/av8rwm
http://tinyurl.com/chwjms
Mark.
----------------------------------------------------------------------
--
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/juniper-nsp

?