Targeted LDP Sessions Flapping on a Cisco 7600 series router

[davicho182]

Hello everyone,

I’m currently trying to figure out the cause of some flapping LDP sessions that i have running on a PE device which is a 7600 series router. There are currently 8 networks that have this issue. These networks are not directly connected to the PE so we are using target LDP sessions. When using the show logging command there are no signs of the LDP sessions dropping out, but if we use the command Show mpls ldp neighbor | i Peer|Up we can see that these LDP sessions have gone up recently

Example:

XXE01#show mpls ldp neighbor | i Peer|Up

Peer LDP Ident: X.X.X.A:0; Local LDP Ident X.X.X.Y:0

    Up time: 00:26:05

    Peer LDP Ident: X.X.X.B:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:25:53

    Peer LDP Ident: X.X.X.C:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:25:45

    Peer LDP Ident: X.X.X.D:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:25:42

    Peer LDP Ident: X.X.X.E:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:25:42

    Peer LDP Ident: X.X.X.F:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:23:49

   Peer LDP Ident: X.X.X.G:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:23:18

    Peer LDP Ident: X.X.X.H:0; Local LDP Ident X.X.X.Y:0

        Up time: 00:23:07

I have omitted the ip addresses and the hostname for obvious reasons.

I would like to know if anyone has gone through a similar scenario that could give me a hint of what could be causing these sessions to restart randomly.

[davicho182]

I would like to point out that these random restarts of the targeted ldp sessions are also affecting the l2vpn’s that are using these target ldp’s.

XXE01#show mpls l2transport vc YYYY detail   

Local interface: VlXXX up, line protocol up, Eth VLAN XXX up

   MPLS VC type is Eth VLAN, interworking type is Ethernet

   Destination address: A.B.E.F, VC ID: YYYY, VC status: up

     Output interface: TeX/X/0, imposed label stack {xxxx yyy}

     Preferred path: not configured

     Default path: active

     Next hop: A.B.C.D

  Create time: 8w3d, last status change time: 00:01:01

   Signaling protocol: LDP, peer A.B.E.F:0 up

     Targeted Hello: A.B.E.H(LDP Id) -> A.B.E.F

     Status TLV support (local/remote)   : enabled/supported

       Label/status state machine        : established, LruRru

       Last local dataplane   status rcvd: no fault

       Last local SSS circuit status rcvd: no fault

       Last local SSS circuit status sent: no fault

       Last local  LDP TLV    status sent: no fault

       Last remote LDP TLV    status rcvd: no fault

     MPLS VC labels: local zzzz, remote yyy

     Group ID: local 0, remote 0

     MTU: local 1900, remote 1900

     Remote interface description: PPPOE_XXE01

     MAC Withdraw: sent:8688, received:0

   Sequencing: receive disabled, send disabled

   SSO Descriptor: A.B.E.F/YYYY, local label: xxx

     SSM segment/switch IDs: xxxxx/xxxx (used), PWID:xxxxxx

   VC statistics:

     packet totals: receive 253111220373, send 172365557653

     byte totals:   receive 293736234082631, send 29184289664599

     packet drops:  receive 0, seq error 0, send 0

[Peter Paluch]

Hello David,

This is a blind shot and perhaps I am asking about an obvious thing - but have you actually had a look in your router's logs? Every LDP neighbor's up or down transition should be recorded with a terse explanation why the neighbor went down. I suggest starting at this point.

Best regards,

Peter

[davicho182]

Hello Peter,

Yes i’ve checked the logs on the PE that’s reporting the issue and there isn’t anything about dropping an LDP neighbor. I’ve been tshooting one specific network out of the 8 networks that are presenting this problem. Let me show you a segment of the topology we have at the moment:

         PE_A-----------------> P -----------> PE_B

The following ip addresses are configured on loopback “A” on each device:

P: has the ip address W.W.W.W

PE_A: has the ip address X.X.X.X

PE_B: has the ip address Y.Y.Y.Y

LDP sessions are formed between loopback A of each device.

“PE_A” has a target LDP session with “PE_B”. Target LDP session between these 2 devices is restarting randomly. This is the network I’ve been focusing on.

PE_A#show mpls ldp neighbor Y.Y.Y.Y -------> Ip address of neighbor “PE_B” (The one that’s dropping out randomly)

    Peer LDP Ident: Y.Y.Y.Y:0; Local LDP Ident X.X.X.X:0

        TCP connection: Y.Y.Y.Y.11111 - X.X.X.X.646          (well-known ldp port)  

        State: Oper; Msgs sent/rcvd: 1607/1607; Downstream

        Up time: 00:41:13 (Notice the Up time)

        LDP discovery sources:

          Targeted Hello X.X.X.X -> Y.Y.Y.Y., active, passive

Couple of hours later I use the same command again on PE_A

PE_A#show mpls ldp neighbor Y.Y.Y.Y -------> Ip address of neighbor “PE_B” (The one that’s dropping out randomly)

    Peer LDP Ident: Y.Y.Y.Y:0; Local LDP Ident X.X.X.X:0

        TCP connection: Y.Y.Y.Y. 22222 - X.X.X.X.646           (Notice that the tcp port changes when the ldp session resets)

        State: Oper; Msgs sent/rcvd: 1607/1607; Downstream

        Up time: 00:25:24

        LDP discovery sources:

          Targeted Hello X.X.X.X -> Y.Y.Y.Y., active, passive

I’ve had issues with ldp sessions dropping out on other devices before but I have found that in those particularly cases the dropping out of ldp sessions was originated because of MTU limitations on the Tx media. This was solved adjusting the mss (maximum segment size) with the command ip tcp adjust-mss. Also, in that device, there were logs showing the flapping ldp sessions.

Now on this device “PE_A”  I don’t see any logs showing that ldp sessions are flapping, nor on "PE_B".

What I have noticed is that the tcp session also restarts (tcp port numbers changing after ldp session restart).

I think that if I found out why tcp connections are resetting I might be able to find a solution for this issue. Any thoughts?

Regards,

David