A9K-MOD400-TR Insufficient fabric capacity with RSP440

[configcheck1]

Hi,

We're in a bit of a problem right now, we have a MOD400-TR + MPA-2x100G and when we plug into the ASR9006-V2 or ASR9010-V2 chassis throw the following warning:

%FABRIC-FIA-1-RATE_LIMITER_ON : Set|fialc[163908]|0x1073000|Insufficient fabric capacity for card types in use - FIA egress rate limiter applied 

And of course if i check the throughput on the MPA ports the RX is arounf 100G but the TX is around 39G.

RP/0/RSP0/CPU0:ios#show ver bri
Tue Feb 19 11:50:38.442 UTC

Cisco IOS XR Software, Version 6.4.2[Default]
Copyright (c) 2018 by Cisco Systems, Inc.

ROM: System Bootstrap, Version 0.76(c) 1994-2012 by Cisco Systems, Inc.

ios uptime is 16 minutes
System image file is "disk0:asr9k-os-mbi-6.4.2/0x100305/mbiasr9k-rsp3.vm"

cisco ASR9K Series (Intel 686 F6M14S4) processor with 12582912K bytes of memory.
Intel 686 F6M14S4 processor at 2128MHz, Revision 2.174
ASR 9006 4 Line Card Slot Chassis with V2 AC PEM

1 FastEthernet
2 Management Ethernet
2 HundredGigE
2 DWDM controller(s)
503k bytes of non-volatile configuration memory.
6271M bytes of hard disk.
12510192k bytes of disk0: (Sector size 512 bytes).
12510192k bytes of disk1: (Sector size 512 bytes).
RP/0/RSP0/CPU0:ios# show diag sum
Tue Feb 19 11:50:42.794 UTC

NODE module 0/RSP0/CPU0 ASR9K Route Switch Processor with 440G/slot Fabric and 12GB

NODE module 0/1/CPU0 400G Modular Linecard, Packet Transport Optimized

MPA 0/1/0 : ASR 9000 2-port 100GE Modular Port Adapter

RP/0/RSP0/CPU0:ios#

Could anyone guide me in how to solve this, i did the same exactly configuration for another customer not so long ago and was working fine (so 100G TX/100G RX).

Regards,

[tkarnani]

Do you only have a single RSP?

[configcheck1]

Hi Tkarnani,

First test (the one for the other customer that was working) was only with one RSP440 but this one i tested with one and two SUPs.

I forgot to mention in my first message that i was using a really basic L2 config (ports configured as l2transport and assigned to the same bridge domain).

Right now i have it working in basic L3 but would like to know why in the past it worked in L2 with only one supervisor.

Regards,

[tkarnani]

There are quite a few variables here.

1st , with 1 RSP the fabric will be limited to 220G

2nd, we do not know the dest port, if its a 10G line card for example there is a limitation of VOQ size, which will only allow 13-18G per flow (from 100G LC to 10G LC)

3rd, we do not know the packet size, if it is small size packet for example there is a chance the NP is being overwhelmed and experiencing tail drops

let me know if you have any additional questions

Thanks!

[configcheck1]

Hi,

I understand the fist point, last test was with 2 RSP.

2) Traffic flows from 1 port of the 2x100G to the other (just a required proof for the customer).

3) packet size is 9000 and mtu is configured as 9216 in both interfaces.

Regards,