Cluster problem

[mike.f]

I have 24 switches (mixture of 3524/48 and 2950T/G) in two clusters all hanging off a core catalyst 4006 with SupII running catOS. I am aware of the problem with clustering through this device as it is CDP enabled but does not support clustering so when setting up the clusters the member switches are not auto discovered and i have had to map the mac addresses to the switch names.

However one problem i have is that if a member switch for any reason is rebooted or communication to the cluster commander is lost for a short period the switch is shown as offline in red on the managment screen and will never go back to green even when the switch is back online and working for days.

The only way to rectify the problem i have found is to delete and reinstall the member switch in the command switch IOS. I assume the 4006 is blocking something, has anyone any ideas?

[milan.kulik]

Hi,

I've used following scenario in the past:

Configure an IP address on each member switch

(conf t

int VLANx

ip address ...

management)

You can add the member switch manually via CLI then even when there is an CatOS switch between the management and member switch.

And more, you can Telnet to each member switch in a case of cluster failure.

Regards,

Milan

[gmaindol]

the only workaround is what you did irrespective of cat4K

i have seen it happening numerous times that it shows red even though it works fine . just dismantle the cluster and build it again

[mike.f]

thanks for the replys milan and gmaindol, i already have ip on all the switches which are mapped to mac addresses on the commander, i have 5 or 6 sites that use clustering with no core switch and never have any problems, the only site that experiences this problem is the one with the cat4k.

If anyone reading this that has a similar setup, just out of interest could you reboot one of your members and then see if if is lost on the commander when looking at it with the web based managment, if it works i will know i'm not flogging a dead horse trying to solve this problem.

Rebuilding the cluster is a problem for me as i have a long trip to visit site each time.

[milan.kulik]

Hi Mike,

generally, why do you use Cluster at all, while having an IP address configured on each member switch?

Just to see all the switches installed in the same site on one screen?

If you are using CiscoWorks, you can use CiscoView Stackmaker with the same result and much better reliability.

If not, you have to remain with Cluster Manager.

But you can rebuild your cluster via Telnet connection to switch CLI, not necessary to travel to each site.

Regards,

Milan