High discards after splitting VLANs on core routers

[Josh Morris]

Hello. I have a large L2 network. We previously had all VLAN STP root and HSRP active on one of two core routers. I recently decided to split the STP and HSRP on both switches to help with processor utilization. I went with the even/odd approach.

Things are working well, and no one has complained of any issues, but I am now seeing a very high number of interface discards on one of the two routers. Basically any port connected to a building distribution switch is now take millions of discards every day.

I have verified that STP and HSRP are in fact in the correct states. Is there something I'm missing as to why this would be happening? The is no change in bandwidth after the split. The discards happen throughout the day but are expecially heavy at night during device backups.

I have confirmed that STP on the distribution side is blocking the correct ports. Is it possible for asynchronous L2 switching? As in can packets go down one interface from the core to the distribution then come back up another itnerface to a different core router?

Any ideas? Thanks.

[Peter Paluch · Accepted solution]

Hi Josh,

Perhaps this can be related to asymmetric routing/switching that occurs quite often with split HSRP Active/Standby routers, as described here:

http://www.cisco.com/en/US/tech/tk648/tk362/technologies_tech_note09186a0080094afd.shtml#t8

Do you believe you could try to align the MAC table aging time with the ARP aging time, and see if this alleviates the packet drops?

Best regards,

Peter

[Josh Morris]

Sounds like this is my issue. I will try the document's recommendations. Thanks!

[Peter Paluch]

Josh,

You are welcome - just please let us know if it indeed solved your issue.

Best regards,

Peter

[matthew-frederick]

Do you know if they are tx discrds or rx discards?

[Josh Morris]

They are TX, coming from the switch who holds the root and hsrp active for our server subnet.

[matthew-frederick]

I had a similar issue where one of our main switches was having 100,000,000+ transmit discards and was causing latency on the network. It was flooding it out all distro port. So what doeas switch do with a mac address it doesn't know? Floods it out all ports. I cleared the ARP on the culprit switch and it cleared everything up. Must have been a dynamically learned MAC address or something that got cleared and the issue hasn't come back up since.