6513 port-channels fail on high throughput...

[dmorassut-sbi]

I'm having a problem with port-channels failing between two 6513 chassis with sup 720s and the 6724-SFP line cards in high throughput situations. The IOS is 12.2(18)SFX5. Has anyone else ever encountered such issues? If so, what was the resoution. Thanks! Dennis.

[glen.grant]

could be something like this.

CSCeg49196 Bug Details

Headline Excessive Overruns and lbusDrops due heavy flow control over fabric

Product IOS

Feature OTHERS Components Duplicate of

Severity 1 Severity help Status Resolved Status help

First Found-in Version 12.2(17d)SXB02 All affected versions First Fixed-in Version 8.5(0.21)JAC, 8.4(1.19), 8.6(0.51)TAL, 8.4(1.21), 12.2(17d)SXB07, 8.4(1.43), 8.4(1.45), 8.5(0.44)JAC, 8.6(0.69)TAL, 8.4(0.61)RCC05, 12.2(18)SXD04 Version help

Release Notes

A Sup720 system where the majority of modules are of type WS-X67xx may

experience system traffic flow degradation under certain traffic profiles,

typically meshed communication among hosts connected to all modules in the

system plus many-to-one type of communications. The symptoms seen during this

condition include:

- overruns on almost all active interfaces

- lbusDrops associated with the supervisor slot (seen via "show fabric channel-

counters")

- TestMacNotification test failed messages seen repeatedly via "show log" for

modules with DFC's installed

- ports leaving and re-joining an EtherChannel

- Routing protocol flaps (hsrp, ospf ...)

- line protocol for multiple interfaces going down

Typically the time interval between occurrences of the problem fluctuates

greatly. During the problem overall PPS (packets-per-second) performance is

drastically reduced through the system as seen via the "show mls statistics"

command.

Workaround:

Reduce the amount of traffic flowing through the switch, particularity on

interfaces seeing the most overruns.

[dmorassut-sbi]

Thanks Grant. The issue has to do with system backups across the core causing the port-channels to fail, so this is definitely on the right track. It affects the 6724 and 6748 modules.

This was helpful! Thanks again.

Dennis.

[jkeeffe]

CSCeg49196 Bug Details states it was 'First Fixed-in Version .... 12.2(18)SXD04, but Dennis' version is 12.2(18)SXF5.

Does that mean Dennis' problem may be related to something else?

[glen.grant]

SXF is a different train than sxd, if you look at the affected versions of this problem his IOS version was on the affected list so its possible it could affect him .