ASR1001-X BGP soft-reconfiguration inbound question

[TelesEC]

Hello everyone,

I have a big doubt and I hope you can help me, thanks in advance.

Question: Why is the memory consumed by adj-rib-in not greater than or equal to the memory consumed by local-rib?

Context:

I was reviewing the configuration of my Internet Edge routers and in the BGP configuration with the ISP we have applied the following command:

neighbor x.x.x.x.x soft-reconfiguration inbound

The ISP sends me the full BGP routing table and I am aware that the command 'soft-reconfiguration inboud' should be deleted.

But I don't understand why the memory used by the Adj-rib-in is not the same as the local-rib?

show ip bgp neighbors x.x.x.x.x
*Output omitted for brevity*
Sent Rcvd
Prefix activity: ---- ----
Prefixes Current: 7 818518 (Consumes 111316272 bytes)
Prefixes Total: 7 4120075
Implicit Withdraw: 0 3039791
Explicit Withdraw: 0 261766
Used as bestpath: n/a 777318
Used as multipath: n/a 0
Used as secondary: n/a 0
Saved (soft-reconfig): n/a 1 (Consumes 136 bytes)

show ip bgp summary
*Output omitted for brevity*
1 received paths for inbound soft reconfiguration

PID: ASR1001-X Version 16.9.5

show ip bgp neighbors x.x.x.x policy
Neighbor: x.x.x.x, Address-Family: IPv4 Unicast
Locally configured policies:
prefix-list ISPtoME in
prefix-list ANNONCE_ME_ISP out
filter-list 7 out
soft-reconfiguration inbound
remove-private-as

[MHM Cisco World]

any update form one peer will be save in 
Adj-RIB-IN
after that it will filter with INBOUND policy then it will be add to 
Loc-RIB

at end any other routes are discard from Adj-RIB-IN

with 
soft-reconfig inbound 

the routes are copy from Adj-RIB-IN, 
when we do 
clear ip bgp soft in

the router only refilter the route again with new INBOUND policy if found. 

[TelesEC]

Hello,

Thank you very much for taking the time to answer my question.

What I am looking to understand is why the memory consumed by local-rib is not the same as adj-rib-in.

Normally with this command, I should store twice the prefixes announced by my peer, but according to my router, this is not the case.

Prefixes Current: 7 818518 (Consumes 111316272 bytes)
....

Saved (soft-reconfig): n/a 1 (Consumes 136 bytes)

[TelesEC]

Hello,

Thank you very much for your time and investigations.... I still can't find a logical explanation.

On the other hand, and as far as I know... If I have configured the feature soft-reconfiguration, the BGP route refresh capability is not used, even though the capability is negotiated.

Andrés

[paul driver]

Hello
I would suggest to remove soft-reconfiguration as it’s an old way to refresh your rtrs bgp policy after a change against all your neighbours advertised prefixes they send to you.

Using soft-reconfiguration the bgp rtr has to store each neighbouring bgp peers unmodified routes so they can be checked against your new policy before they are actually placed the bgp route table (soft reset) and this takes up rtr resource such as memory and if you had multiple bgp peerings will large rib tables that’s a lot or resource being used.

Most rtrs should support route-refresh which dynamically requests updated route information from it peers (dynamic soft-reset) thus negates the need to store any peer information locally.

To check if you bgp rtr and it peers support route-refresh

show ip bgp neighbours | in refresh

[TelesEC]

Hello Paul.

Thank you very much for your reply.

I edited the question a bit to be clearer.

I am aware that the command should be suppressed, what I don't understand is why the memory used by the adj-rib-in is only a few bytes since I receive the whole internet table.

Because as I see in my router, having this command configured, does not affect the router's memory... which is normally not true.

Regards,

Andrés

[paul driver]

Hello

Thank you for the clarification

---

@TelesEC wrote:

Question: Why is the memory consumed by adj-rib-in not greater than or equal to the memory consumed by local-rib?

---

My understanding here is that the local rib prefixes are those that have been processed against you inbound bgp filter policy and are deemed viable to be entered into the bgp table any other prefixes left the adj-rib-in are left until they become resolvable or they are withdrawn ,However when using soft-inbound configuration a copy of that adj-rib-in is also separately stored.

[TelesEC]

From what I understand, when configuring the 'soft-reconfiguration inboud' command.
A memory space is dedicated to store the prefixes before applying any policy.

In this case the ISP announces me the full BGP table... so my router, inefficiently, should store twice the entire Internet table, once in the local-rib and the other in the adj-rib-in... but according to the output this is not happening.

Prefixes Current: 7 818518 (Consumes 111316272 bytes)
....

Saved (soft-reconfig): n/a 1 (Consumes 136 bytes)

For example on another router, a C6XXX,  I get this behavior.  I receive 1215 prefixes and only accept 187.

 Sent Rcvd
Prefix activity: ---- ----
Prefixes Current: 58 187 (Consumes 117768 bytes)
...
Saved (soft-reconfig): n/a 1215 (Consumes 102060 bytes)

[Georg Pauwen]

Hello,

I don't really have a direct answer, but actually, that is how I always understood it as well: soft reconfiguration actually doubles the memory used. As far as I recall, there are two parts to the adj-RIB, the adj-RIB-in, and the adj-RIB-out. The 'in' contains all the unedited information, while the 'out' contains the information after outbound polciies are applied, Not sure if your specific router has those policies configured ? Either way, only the adj-RIB-in table is stored in memory for each neighbor, the adj-RIB-out table not.

So I wonder what the output:

--> Saved (soft-reconfig): n/a 1 (Consumes 136 bytes)

is actually referring to.

[TelesEC]

Hello Georg,

Thank you for your response.

I don't understand the behavior on these routers (ASR1K)

On another router, a catalyst 68XX, the situation is different and the behavior is as expected.

 Sent Rcvd
Prefix activity: ---- ----
Prefixes Current: 58 187 (Consumes 117768 bytes)
Prefixes Total: 489 24837
Implicit Withdraw: 405 12267
Explicit Withdraw: 26 12382
Used as bestpath: n/a 187
Used as multipath: n/a 0
Used as secondary: n/a 0
Saved (soft-reconfig): n/a 1215 (Consumes 102060 bytes)

In other words, I accept 187 prefixes but receive 1215 prefixes.
I don't know why in Internet Edge routers I don't have the same behavior.

I don't want to be left with that doubt

Regards,

Andrés

[MHM Cisco World]

If i am right this is edge router receive few or only one route which is defualt, but it adverise many route toward SP.

Keep in mind Loc-RIB have the IN and OUT route.

[TelesEC]

No, my router receives the full BGP routing table

You can see it in the following output, 818528 prefixes:

show ip bgp neighbors x.x.x.x.x
*Output omitted for brevity*
Sent Rcvd
Prefix activity: ---- ----
Prefixes Current: 7 818518 (Consumes 111316272 bytes)
....
Saved (soft-reconfig): n/a 1 (Consumes 136 bytes)

 Section "Saved (soft-reconfig):"  should have about the same number of routes.

I don't understand why the difference.
When I use the command.. show ip bgp neig x.x.x.x.x received-routes... I get the whole internet table.

Andrés

[MHM Cisco World]

sh ip bgp neighbors x.x.x.x | s Neighbor capabilities

please can you show the output for both ASR1000 and other C6xxx

[TelesEC]

Hello,

Here you have the requested information.
For your information, the ASR is the Internet Edge, the C8XX is a PE.

For ASR1001-X

 Neighbor capabilities:
Route refresh: advertised and received(new)
Four-octets ASN Capability: advertised and received
Address family IPv4 Unicast: advertised and received
Graceful Restart Capability: received
Remote Restart timer is 120 seconds
Address families advertised by peer:
IPv4 Unicast (was preserved, Peer capable of Enhanced GR)
Long-lived Graceful Restart Capability:
Enhanced Refresh Capability: advertised
Multisession Capability: 
Stateful switchover support enabled: NO for session 1

A few lines after this section we have.

Inbound soft reconfiguration allowed

For C68XX

 Neighbor capabilities:
Route refresh: advertised and received(new)
Four-octets ASN Capability: advertised and received
Address family IPv4 Unicast: advertised and received
Enhanced Refresh Capability: advertised and received
Multisession Capability: 
Stateful switchover support enabled: NO for session 1
....
Inbound soft reconfiguration allowed

 Thanks

[MHM Cisco World]

Can you give some time, 
this interest, 
What I suspect ?
there is 
A-soft-reconfig "save a route in DB and re process it when do clear ip bgp soft"
B-route refresh "automatically ask the neighbor a route"

BUT there is 
Soft Reset Enhance <- I know it new for me also 

this reduce the memory need for save a route in DB but the full technology I don't get until now so I will do more search this week and I will update you.