mac-address-table not updating

[nathan.driver]

I have a problem with the mac-address table not updating on a 6509, running IOS (s222_rp-IPSERVICESK9-M), Version 12.2(18)SXF16

I have two 6500's connected together and the other one is running code, Version 12.2(18)SXF17a does not have the issue.

To replicate my problem i have a VMware setup with host servers connected to both switches, if i Vmotion a VM from one host on Switch A port X to another host on Switch A port Y the mac-address-table does not update and therefore the VM is un-contactable over the network.  To remedy the problem i manually clear the mac address for the VM on Switch A.

Now if i do this same test with the same host servers on Switch B the mac-address-table updates in milliseconds and the VM never has any network downtime.

Also if i move a VM between hosts that are on different Switches i also get the problem.  Its basically any change to a mac address on Switch A causes the VM to go network offline.

I have also had the issue with physical connected Windows servers when we physically move their network connection.  I only mention VMware because I can replicate the problem from my desk.

I have looked at the Mac address aging time and its the default 5 minutes on both switches.  I have read that the arp table timeout value could be a factor but i have yet to find the command to see the set value.

Their is of course a minor code difference so maybe this could be an issue, but I to get the outage to fix this is really tough so I can't do it unless i'm really confident it will fix it.

If anyone has any suggestions how i could resolve the problem it would be appreciated.

Many thanks in advance

Nathan 

[Shashank Singh]

Hi Nathan,

List of all the bugs resolved between these two IOS versions are available at

http://www.cisco.com.az/en/US/docs/switches/lan/catalyst6500/ios/12.2SXF/native/release/notes/OL_4164.pdf

However, I do not see any bug matching your symptom.

To check the ARP timeout value do the following command:

Switch>sh int gi1/0/1

GigabitEthernet1/0/1 is down, line protocol is down (notconnect)

  MTU 1500 bytes, BW 10000 Kbit, DLY 1000 usec,

     reliability 255/255, txload 1/255, rxload 1/255

  Encapsulation ARPA, loopback not set

  Keepalive set (10 sec)

  Auto-duplex, Auto-speed, media type is 10/100/1000BaseTX

  input flow-control is off, output flow-control is unsupported

  ARP type: ARPA, ARP Timeout 04:00:00   <<---------

Regards,

Shashank

Please rate if you found this post helpful

[nathan.driver]

thanks for getting back to me shashank

The arp timeout values are set to the default 4 hours, so i dont think that should be causing the problem.

I have had a browse of those pdf's but like you i cant find my issue listed.

Does anyone have any idea what logging level i need on the switches and the info i might look for on the syslog to make sure when i move a VM the physical switch is receiving the packet that says update mac-address table?

thanks

Nathan

[nathan.driver]

I have worked out what this is.

During the last 6 months we have had several sup2's replaced and i was told they were identical to what i had installed.  However as i have now discovered this is not quite true.

This is what i did have

WS-X6K-SUP2-2GE

This is what they replaced it with

WS-X6K-S2-MSFC2

The repalcements has HW version 6.1 whilst the original has 3.2

So on one switch my active supervisor was the newer replacement and on the second switch was the original.  It was this that was causing my mac-address update issue.

At the weekend i therefore failed over the supervisors on the second switch to make sure the active supservisors were both the newer replacements running the 6.1 hardware.

As soon as i did this the mac issue had gone which was good news.

I am  going to get the remaining supervisors running the old hardware swapped out, to make sure I have all the same HW versions.

many thanks

Nathan