12-15-2012 08:29 AM
Hi. everyone
I'm trying to configure bgp accounting on interface bundle-ether on ASR 9010 (IOS-XR:4.2.0),
But command
sh cef ipv4 interface bundle-ether1 bgp-policy-statistics
show zero counts of bytes and packets.
I looking for answer on question: It is hardware or software restriction?
Can experts say, bgp accounting on interface bundle-ether must work or not.
Not trying yet IOS-XR: 4.2.3. I'll updrade my working ASR to 4.2.3 if that helps to resolve issue.
Using bgp accounting for counting traffic on TenGigabitEthernet interfaces and it"s works fine.
Tnanks in advance.
Solved! Go to Solution.
12-20-2012 12:31 PM
You description is the perfect match for CSCtr13320 41126I : IP-IP_EA-3-PLATFORM_UPDATE_BULK Not enough memory msg flood
Symptom:
Repeated "IP-IP_EA-3-PLATFORM_UPDATE_BULK Not enough memory" messages are flooded on the console, with no specific user trigger.
Condition:
The issue occurs only when BGP-PA is configured on bundle-ether interfaces and in certain scenarios,
usually seen when all the members of the doesn't span across all NP's of the LC.
In case of bundle, PM tries to allocate required BGP-PA stats for all NP’s, eventhough there is no member on that NP, which is not correct. Hence this
error msg is displayed.
Workaround:
Remove BGP-PA config from the Bundle (or) have members of Bundle come from all NPs on the Linecard.
From what I see the bug should have been fixed since 4.2.0. Can you confirm you run 4.2.0?
You’ve mentioned you wanted to try 4.2.3, may you do so?
Regards,
/A
12-16-2012 11:58 AM
Hello Dmitriy,
Yes, it should work. May you try using location keyword to specify the bundle member location?
show cef ipv4 interface bundle-eth1 bgp-policy-statistics location x/y/CPU0
Regards,
/A
12-20-2012 12:56 AM
Thanks for answer. I did not exactly expressed my issue. Output non zero, but we see very small number of counted bytes and packets. Bytes in second where traffic is more than 10Gbit/s.
Output of commands with bundle member location and without are identical:
RP/0/RSP1/CPU0:cisco#sh cef ipv4 interface bundle-ether 1.800 bgp-policy-statistics location 0/1/CPU0
Thu Dec 20 14:17:44.821 ESS
Bundle-Ether1.800 is UP
Input BGP policy accounting on dst IP address enabled
buckets packets bytes
default 2 88
1 1 28
3 2 64
4 1 44
5 4 128
6 72 4082
63 1 44
Input BGP policy accounting on src IP address enabled
buckets packets bytes
default 25 2400
63 89 4766
Output BGP policy accounting on dst IP address enabled
buckets packets bytes
Output BGP policy accounting on src IP address enabled
buckets packets bytes
RP/0/RSP1/CPU0:cisco#sh cef ipv4 interface bundle-ether 1.800 bgp-policy-statistics
Thu Dec 20 14:17:46.301 ESS
Bundle-Ether1.800 is UP
Input BGP policy accounting on dst IP address enabled
buckets packets bytes
default 2 88
1 1 28
3 2 64
4 1 44
5 4 128
6 72 4082
63 1 44
Input BGP policy accounting on src IP address enabled
buckets packets bytes
default 25 2400
63 89 4766
Output BGP policy accounting on dst IP address enabled
buckets packets bytes
Output BGP policy accounting on src IP address enabled
buckets packets bytes
I noticed in log repeating message, when enabled bgp accountig on bundle-ether.
LC/0/1/CPU0:Dec 20 13:52:49.084 : ipv4_io[228]: %IP-IP_EA-3-PLATFORM_UPDATE_BULK : Platform update failed for 1 interface: FIB_FEATURE_IF_TYPE_IPV4_ALL_INTF_INFO - Not enough memory
After disabling bgp accountig on bundle-ether, message disappeared.
Not noticed this messages with enabled bgp accountig on routed TenGigabitInterface. Our processor A9K-RSP-4G and card A9K-8T-L. What restriction in memory we achieved? It's very strange, because after enabling bgp accountig on bundle-ether and getting issue, we enabled bgp accountig on other subinterfaces on routed TenGigabit ports and accountig works fine there.
12-20-2012 12:31 PM
You description is the perfect match for CSCtr13320 41126I : IP-IP_EA-3-PLATFORM_UPDATE_BULK Not enough memory msg flood
Symptom:
Repeated "IP-IP_EA-3-PLATFORM_UPDATE_BULK Not enough memory" messages are flooded on the console, with no specific user trigger.
Condition:
The issue occurs only when BGP-PA is configured on bundle-ether interfaces and in certain scenarios,
usually seen when all the members of the doesn't span across all NP's of the LC.
In case of bundle, PM tries to allocate required BGP-PA stats for all NP’s, eventhough there is no member on that NP, which is not correct. Hence this
error msg is displayed.
Workaround:
Remove BGP-PA config from the Bundle (or) have members of Bundle come from all NPs on the Linecard.
From what I see the bug should have been fixed since 4.2.0. Can you confirm you run 4.2.0?
You’ve mentioned you wanted to try 4.2.3, may you do so?
Regards,
/A
12-21-2012 11:26 AM
Thanks, seems it is what we are looking for.
Running version:
RP/0/RSP1/CPU0:cisco#sh version brief
Cisco IOS XR Software, Version 4.2.0[Default]
RP/0/RSP1/CPU0:cisco#sh install active summary
Active Packages:
disk0:asr9k-mini-p-4.2.0
disk0:asr9k-k9sec-p-4.2.0
disk0:asr9k-mgbl-p-4.2.0
Fixed in: 4.2(0.14)I
As I guess, we can fix bug without interrupting traffic via ASR9010.
But I can't find correct pie between files in
IOS XR Software Maintenance Upgrades (SMU)-4.2.0 and in
4.2.0-Updated Tarball for ASR9K Recommended SMU's: 4.2.0_asr9k-p_REC_SMUS_2012-10-04.tar
Please, say the exact name of 4.2.0 pie, that fix bug.
It will be the best choice, that help us - don't interrupt traffic via our ASR9010.
12-23-2012 01:42 PM
Hello Dmitriy,
As I’ve mentioned the bug should have been fixed already in 4.2.0 but apparently it was not.
Open a TAC Service Request. We’ll check the fix commitment with the dev team.
Regards,
/A
12-24-2012 12:53 AM
Thanks, Alexei. I opened TAC Service Request. I hope to get SMU.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide