2900/3500 Clustering & NMS Alarms: Opinions Wanted

[jwitherell]

I have tried to use the cluster management features of the 2900 and 3500 switches. But I stopped using it, as my network monitoring station couldn't ping each one. So now, in my stacks of 3524s (for example), I configure each one separately, with their own IP.

Am I taking the easy way out? For the folks who use Cluster Manager, how do you know when one of the switches fail? Do all of you use Cisco Works? Does anyone use clustering and have a multiplatform NMS, like NetView, OpenView or WhatsUp Gold? How do you ensure that someone in the NOC knows when the switch goes down? Especially off-hours?

This may be something good for alot of people. It would be great to hear how different people handle this.

[mmoscoso]

When you use a cluster all the switches respond to a unique IP address. If you are not using Cisco Works you may think in monitor not the ip address of a cluster member but the mac address.

Other more practice solution could be monitor the ip address of one device attached to the switch ( a PC for example) so you'll know when the switch goes down.

Hope this help.

Mauricio

[milan.kulik]

It is possible to give an IP address to each switch in the cluster. This is the way I administer my network and everything is OK.

HTH,

Milan

[jwitherell]

I guess up till this point you either had a choice of using clustering with the single IP address, or *not* using clustering and give each one an IP address.

I'll have to try that. So you you assign the address and then add it to the cluster, or add the IP after adding it to the cluster?

[BRUNO WOLLMANN]

I have many 3500XL switches on my network and use HPOV NNM. I used to run my switches in cluster mode but was unhappy that NNM could not manage each switch independently. I was also unhappy with the web interface (I prefer the CLI). I am also using CiscoWorks 2000 and it was unable to manage each switch individually while they were in cluster mode. I experimented with assigning IP addresses to each switch while running in cluster mode but stopped doing that for a few reasons. I didn't want the CMS traffic running between my clustered switches (even though the traffic is probably minimal). I like using the Configuration Management function in CW2000 because it allows me to type in the IOS commands and then ship them out to a group of switches all at once. I perform the change on a test switch first, if it passes, I send the change to the rest of my switches.

If you don't have CW2000, you may still want to cluster switches so that you can make changes 1 cluster at a time but you can still assign individual IP addresses so that your NMS can monitor each switch.

[milan.kulik]

I assign IP address first and then add the switch to the cluster. I'm not using CW2000 so far, so I don't know if it makes any problem to it. But there is no problem with Cluster Manager.

Regards,

Milan