cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
675
Views
1
Helpful
3
Replies

Best Practice for Swapping EOL Leaf model with New Leaf Model

vv0bbLeS
Level 1
Level 1

Hello all,

I have some older Leaf Models (N9K-C93180YC-EX) that are End-of-Life (EOL) that we will be replacing in the next year or two, and I wanted to start planning on what procedure I will use to swap these out with the newer Leaf model (N9K-C93180YC-FX3).

 

I have found documentation that talks about replacing a failed leaf with an RMA, but in my case the existing Leaf is working just fine and has live production hosts on it, and I would just need to replace it with a newer Leaf.

 

My fabric has NodeID's of 201 - 235 in use right now. So, let's say I'm replacing my EOL Leaf 201 - My initial plan would be:

  1. Provision a new Leaf with a new NodeID (say 251). This would include all initial Leaf setup stuff like Upgrade/Downgrade version, OOB mgmt, Switch/Interface Profiles, etc.
  2. Setup the ports on the new Leaf just like the existing Leaf 201 ports are (same Port Policies, Port Selectors, EPG's, etc.)
  3. Then move host connections from EOL Leaf 201 to new Leaf 251 (hosts would take a short outage of a few seconds while cable is moved).

 

Of course, the above procedure would change a bit if EOL Leaf 201 were part of a VPC domain (say with EOL Leaf 202), in which case I would stand up 2 new Leafs (251 and 252) and perform the same steps above with the added step of setting up the VPC domain on the new Leafs and including any VPC's in any Port Selector or port EPG config.

 

I wanted to put this plan out there and ask if there are any best practices for this that I'm missing? I guess you could call this a rough "Request for Comments" ? : )

0xD2A6762E
2 Accepted Solutions

Accepted Solutions

ecsnnsls
Level 1
Level 1

Hi @vv0bbLeS,

You should probably decommission the existing switch and then commission the new switch. If your production hosts are dual homed, you can replace one switch at a time.

https://www.cisco.com/c/en/us/support/docs/cloud-systems-management/application-policy-infrastructure-controller-apic/213617-aci-leaf-or-spine-replacement-procedure.html#anc8

HTH

View solution in original post

Hefe2
Level 1
Level 1

Hi,

is there a possibility you can dual home your hosts to a 2nd (new) leaf? Then I´d suggest to install a new leaf switch (e.g node 251) and put it in a vpc domain with node 201.

Since you`re going to have an outage anyway  I´d "decommission or "remove from controller" leaf 201, commission the new switch as node 201 and move host connections to the new switch at the same time. This way may cause longer outage, but you don´t need to prepare any config or don´t have any parameters (static bindings,etc) missing on the new switch. 

Greetz

View solution in original post

3 Replies 3

ecsnnsls
Level 1
Level 1

Hi @vv0bbLeS,

You should probably decommission the existing switch and then commission the new switch. If your production hosts are dual homed, you can replace one switch at a time.

https://www.cisco.com/c/en/us/support/docs/cloud-systems-management/application-policy-infrastructure-controller-apic/213617-aci-leaf-or-spine-replacement-procedure.html#anc8

HTH

Hefe2
Level 1
Level 1

Hi,

is there a possibility you can dual home your hosts to a 2nd (new) leaf? Then I´d suggest to install a new leaf switch (e.g node 251) and put it in a vpc domain with node 201.

Since you`re going to have an outage anyway  I´d "decommission or "remove from controller" leaf 201, commission the new switch as node 201 and move host connections to the new switch at the same time. This way may cause longer outage, but you don´t need to prepare any config or don´t have any parameters (static bindings,etc) missing on the new switch. 

Greetz

vv0bbLeS
Level 1
Level 1

@ecsnnsls yes the hosts are dual-homed, but most of them are dual-homed using a VPC connection, which I guess shouldn't matter as when I decom Leaf 201, the VPC connection to Leaf 202 would still be active and the hosts should stay online while the new Leaf 201 is provisioned. Then I could re-connect the hosts to Leaf 201, confirm connectivity is good, and then repeat the process for Leaf 202.

 

@Hefe2 yes it would be nice to keep the same Node ID's for sure!

 

Thank you both for your suggestions! Thankfully we have a lab environment so I will test these methods in the lab and see what works best for us. Thanks again!

0xD2A6762E

Save 25% on Day-2 Operations Add-On License