Description

This article describes the behavior of a misconfigured LFI with the MLPPP on M120.

Symptoms

After migrating the MLPPP circuit with CE from M20 to M120, the connection was lost.

Environment:

  • LS- (Link Service PIC) was used on the M20.
  • LSQ- (AS2 Layer-2 Services PIC) was used on M120.
  • CE is a Cisco router.

Command output:

> show interfaces terse
Interface                      Admin       Link        Proto        Local            Remote
......
ce1-1/2/7                     up            up
ds-1/2/7:1                    up            up
ds-1/2/7:1.0                  up            down       mlppp        lsq-1/3/0.2
.....
lsq-1/3/0                     up            up
lsq-1/3/0.2                   up            down       inet         202.160.194.189/30


> show log messages | match ppp
Apr 9 14:50:30 m120-re0 /kernel: if_pfe_mlppp_ifl_config lsq-1/3/0: ifl config: op: 2; sub: 2; flags: 0x8010, curr: 0xc010. rlsq 0
Apr 9 14:50:30 m120-re0 /kernel: [mlppp] ds-1/2/7:1.0 of lsq-1/3/0.2 is down
Apr 9 14:50:30 m120-re0 /kernel: [mlppp] ds-1/2/7:1.0 leaves lsq-1/3/0.2
Apr 9 14:50:30 m120-re0 /kernel: [mlppp] lsq-1/3/0.2 reset remote MRRU to N/A
Apr 9 14:50:30 m120-re0 /kernel: [mlppp] reset lsq-1/3/0.2 family MTUs to configured value

Solution


After troubleshooting, it was determined that the LFI with the MLPPP on M120 was misconfigured.

The MLPPP configuration of the working M20 was compared with the MLPPP configuration on M120:


MLPPP configuration on M120:

> show configuration interfaces lsq-1/3/0 unit 2
encapsulation multilink-ppp;
mrru 1524;
fragment-threshold 128;
minimum-links 1;
multilink-max-classes 4;
......

> show configuration class-of-service interfaces lsq-1/3/0 unit 2
scheduler-map aaa_00007636BKK_Scheduler-Map;
fragmentation-map aaa-TEST;
shaping-rate 256k;

> show configuration class-of-service fragmentation-maps aaa-TEST
forwarding-class {
   premiumrt {
      fragment-threshold 128;
      multilink-class 2;
   }
   standard {
      fragment-threshold 128;
      multilink-class 0;
   }
   business {
      multilink-class 1;
   }
   nc_premiumnrt {
      fragment-threshold 128;
      multilink-class 3;
   }
}


MLPPP configuration on M20 which was working:

> show configuration interfaces ls-0/0/0 unit 2
encapsulation multilink-ppp;
mrru 1524;
fragment-threshold 128;
interleave-fragments;

"interleave-fragments" is NOT valid on LSQ-,  to remove MCML from M120, just enabled LFI as below while keeping the CE configuration unchanged.

> show configuration interfaces lsq-1/3/0 unit 2
encapsulation multilink-ppp;
mrru 1524;
fragment-threshold 128;
minimum-links 1;              # “multilink-max-classes” was removed
......

> show configuration class-of-service interfaces lsq-1/3/0 unit 2
scheduler-map aaa_00007636BKK_Scheduler-Map;
fragmentation-map aaa-TEST;
shaping-rate 256k;

> show configuration class-of-service fragmentation-maps aaa-TEST
forwarding-class {
   premiumrt {
      no-fragmentation;   # “multilink-class” option was set to use "no-fragmentation" option to make the voice traffic in this queue become LFI packets, while keep the packets in other queue to be fragmented into 128 Bytes
   }
}

Conclusion:
When comparing, it was concluded that the CE was using LFI with MLPPP, NOT Multi-class MLPPP.  Whereas customer enabled MCML on the M120, which caused the LCP negotiation between M120 and CE to fail.