03-07-2012 06:55 AM - edited 03-07-2019 05:24 AM
When replacing a chassis a couple nights ago, we ran into a strange problem.
Here's a simplified version of the network:
We had to swap out a 6509 with a 6509-E chassis (Core Switch 2).
When Core Switch 2 came back online, we checked CDP and Trunking and all seemed well, but soon got a call that some of the servers were unavailable.
The "unavailable" servers were connected to IDF Switch 1. I tried to telnet to Switch 1, but could not.
I checked the arp table on Core Switch 1 and it contained an entry for IDF Switch 1.
I checked the arp table on Core Switch 2 and it had an incomplete arp entry for IDF Switch 1.
I consoled into IDF Switch 1 and cleared its arp table for vlan 359 (snmp/management) and the switch came back online.
However, the servers (vlan 600) were still unreachable.
When I bounced the links between the Core and the IDF, everything went back to normal.
I cannot replicate this issue in the lab, so I can't determine why it happened. Anyone ever see something like this before?
The only things that were different between IDF switch 1 & 2 is:
IDF 1 - 4948 / vtp transparent
IDF 2 - 4513 / vtp client
Equipment:
Core Switch 1 & 2: 6509 with Sup720 running 12.2(18)SXF4
IDF Switch 1: 4948 running 12.2(25)EWA9.
Thanks!
Ven
03-07-2012 07:11 AM
Hi Ven,
Are you, by any chance, running VTP Pruning?
Best regards,
Peter
03-07-2012 07:14 AM
Peter: Pruning mode is enabled.
Ven
03-07-2012 08:33 AM
Hello Ven,
Alright, that could theoretically explain these issues.
The thing is that the VTP Pruning sometimes likes to declare all VLANs pruned on a trunk. VTP Pruning functionality is provided by switches sending VTP Join messages every 6 seconds, and each Join message contains a list of used/unused VLANs as seen by the sending switch. The funny thing is that if the VTP Join messages have been arriving on a trunk port and suddenly stop arriving, the receiving switch declares all VLANs as unused and prunes them off the trunk.
This can be very unpleasant if, for example, you migrate a VTP Server/Client switch to a VTP Transparent switch: it will stop originating VTP Join messages and in 6 seconds or so, all surrounding switches will prune all VLANs on their trunks to this switch. I am wondering if some situation like this took place in your network - that without an interface flap, some VTP Joins stopped arriving.
Best regards,
Peter
03-07-2012 11:33 AM
Nothing really happened like that. We only replaced (reloaded) our HSRP secondary core switch. The IDF that was having trouble was always vtp transparent. We've done this several times in the past (at multiple sites) and it only seems to happen at this site. The real problem is that I can't replicate it in the lab. I have all the same equipment, IOS, configs, etc and can't make it happen.
Ven
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