Removing Secondary Interconnect PCIe

[victorcharlesedwards]

Besides shutting down the filers and physically moving the fiber cable, what other steps need to be taken to safely move our failover loop from slot 2 port 4 into slot 1 port 3?

We have two filers (FilerA and FilerB) running DOT 8.0.2 7-Mode as a cluster.  Each filer manages one loop, as well as the failover loop for the other filer's shelves.  We need to free up slot 2 on both filers in order to install an external SAS adapter.

While my gut tells me I can simply shutdown both filers and move the fiber cable from slot 2 port 4 into slot 1 port 3, I wanted to verify with the community if this is the best approach.

Thanks!

[scottgelb · Accepted solution]

With the systems down or in maintenance mode there is no issue changing the source hba port (but can't be done with the system up - see similar discussion a day or two ago on here).  ONTAP will find the disks on any path.  I would go to maintenance mode though and confirm all disks with disk show, storage show -p and aggr status to make sure all disks are available over all paths and all aggrs show up after the change.

[victorcharlesedwards]

Thanks Scott!  Would this be doable in failover?  For example if I:

1. Failover FilerA to FilerB
2. Power-off FilerA
3. Move Cable Over on FilerA
4. Power-on FilerA
5. Run (via maintenance mode) disk show / storage show -p / aggr status
6. Repeat for FilerB

[scottgelb]

Looks good.. Or move cables in maint mode too.. Then cf giveback and repeat. It would be worth opening a case just to have one proactive in case of any issue then close after.

Typos Sent on Blackberry Wireless

[aborzenkov]

There was discussion on toasters about shelf unplug and the answer (from NetApp employee) was - the only safe way is with both heads down. It may appear to work with takeover/giveback but you are still risking panic after days or weeks.

Now, moving shelf connection to another adapter basically means "unplugging" all shelves on old one. So I am not sure. I would simply switch off to be on safe side.

[scottgelb]

Support approved takeover/giveback for us…as long as the node changing the source port is the one taken over. It is a CP issue with NVRAM flush and the diskid…in takeover the down node doesn’t have the issue.