Single FI migration to a cluster

[Krzysztof Zaleski]

Hello,

I have a customer who uses UCS B-series infrastructure with single fabric interconnect (don't ask why). There is UCSM 1.4 running on it. Now, the customer purchased additional FI to create a cluster, as recommended. Now, the second FI arrived with 2.0 code. My question is, if I promote that single FI to a cluster and connect the seconf FI, will it be downgraded to 1.4 by the master, or newer version will take over, although it is on a slave? Should I perform manual downgrade to 1.4 on that new FI before I connect it to existing master? We do not need an upgrade to 2.0 now, 1.4 is doing just fine for this environment, so we want to stick wit it.

Cheers,

Krzysztof

[Robert Burns]

Krysztof,

Never introduce a cluster member FI at a different version.

Best approach would be to downgrade the new FI to the same version first.  Next promote the Standalone FI to primary, then introduce the secondary.  (Though this process should not be disruptive, I would schedule downtime regardless to be safe.

Primary/Secondary roles are not chosen by the higher FW level. 

If you have no need for the 2.0 features or bug fixes, then stay put where you are.  If it's not broken, don't fix it.

Regards,

Robert

[rewilkin]

Very well said Rob and I agree and back your statement 100%

Sent from Cisco Technical Support iPad App

[Krzysztof Zaleski]

Hello,

We have just completed the migration. It went quite well. Just one thing was interesting. When I uploaded new package and requested version change in UCMS (for both: UCSM and kernel), the UCSM itself got restarted and new version became active. However, when I restarted the FI by power off/on the kernel and kickstart were still running previous version. The second reboot activated new NXOS. It happened with two FIs which we were downgrading. Did I miss something or it's just normal behavior?

Cheers,

Krzysztof

[Robert Burns]

Expected behavior.  The proper procedure is to activate just UCSM first, let it activate the image (which will kill any existing UCSM instances), then activate the IOMs (startup version only), and finally activate the FI's one at a time. 

Sounds like you doubled up a step which would schedule the second reboot to activate all endpoints.

Regards,

Robert

[anil.gumadavalli]

Hi there,

If you detailed steps involved in this process, please send me.

I am in the process of compiling a plan to do this. I am just little nervous with all the changes.

thank you,

AK G