Storage Failover with Aggregated Gig Ethernet ports

clo3056 · ‎07-16-2008

We are installing a Network Appliance FAS2050 which has two controllers with two ethernet ports each. The controllers are designed to do fail-over if one of them crashes. They are connected to our CISCO 6509 running IOS 12.1(20)E3. We have successfully aggregated the channels using the following setup:

1. Physical Attachments

A. controller 1

1. port e0a to sw1 int gig7/1

2. port e0b to sw1 int gig7/2

B. controller 2

1. port e0a to sw1 int gig7/3

2. port e0b to sw1 int gig7/4

2. Set up Cisco 6509 Interfaces

A. Create port-channels

1. conf t

2. int port-channel 2

3. ..switchport

4. ..switchport mode access

5. int port-channel 3

6. ..switchport

7. ..switchport mode access

B. Aggregate interfaces

1. conf t (if not already)

2. int gig7/1

3. ..channel-group 3 mode on

4. int gig7/2

5. ..channel-group 3 mode on

6. int gig7/3

7. ..channel-group 2 mode on

8. int gig7/4

9. ..channel-group 2 mode on

This all works with the controllers. The problem is that when one of the controllers is failed, there is no path back to the controller when the other controller address is assumed. Does this involve setting up a Link Bundle? If so, how should this be set up? If not, what is the correct procedure? I have not done this before and I thought perhaps someone in this forum would be familiar with storage system fail-over setups on Cisco switches.

As usual, Cisco support says that the settings should be provided by the Vendor. The Vendor says the settings should be provided by Cisco. Can anyone help me?

Chris

inch · ‎07-16-2008

Hi Chris,

Are you telling us when a storage controller fails or a ethernet controller fails?

Ie: What happens when you pull a single cable from each port-channel?

Cheers

Andrew

clo3056 · ‎07-17-2008

This would be when the storage controller fails. Pulling a single cable does not cause any failure. I did not try pulling both cables of the aggregate.

Someone has suggested that the switch may have the MAC cached in the arp. If a host has the old MAC cached, it pings. If I remove the cached MAC for the storage device from the host, the host will not ping the storage device.

I am dealing with a 6509 with IOS, so the only clear arp option takes all the 450 cached addresses out - not a preferable situation.

For anyone familiar with NetApp equipment, the two aggregated ports are configured as multimode VIFs in dynamic mode using IP Address as the fail-over.

inch · ‎07-17-2008

G'day,

The other head/controller should take over the mac-address from the other head/controller.

Generally there is up to a 60s wait for full failover of a head in a filer anyway.

Cheers

Andrew

sunyu · ‎08-20-2008

The controller should take over all the jobs of the failed controller :ipã€mac etc.

the controller must be in the same vlan.

do not worry anything .

clo3056 · ‎08-21-2008

The problem has been resolved. As you suggested, it was the controller that needed to be set correctly. As it turns out, there was a VIF on each controller that had the same name and it was confusing the fail-over process. It is set up correctly now and works fine.