Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

Hi T, 

Have you considered non-disruptive volume migration instead?

Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

Hello!

Yes, this can be done using what is referred to as Paced CCU or Manual Update.

It is a combination of these two procedures:
IBM Docs: Updating the System manually
and

IBM Docs: Replacing nodes nondisruptively wiht an IBM SAN Volume Controller SV3 node


What this does:

The Hardware replacement procedure prep steps readies your WWPNs and Port Number and makes sure the cluster pre-reqs are met

Then before you actually do the add node commands, you set the -pacedccu flag on the SV3 nodes, this will prevent the SV3 from downgrading it's agent code to match the cluster, it WILL downgrade it's cluster state machine to match the cluster.  This will allow you to run an 8.6.x.x version SV3 in a cluster at 8.5.0.x.

Repeat with every node in the cluster.

The tricky part is that you are reducing the number of IO Groups, so you will need to collapse down/remove the SV1 nodes entirely before the upgrade can be completed, and that may be difficult as you are not meant to go changing configuration during the upgrade process, which means you will need to migrate the IO Groups down first.  It is worth raising a support ticket to get walked through this, as they have been working through some pitfalls for other installs of this type.

I hope that helps.

Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

Hi Evelyn,

how does this work exactly ?

When I execute the following command "applysoftware -file IBM_SVC_INSTALL_8.7.0.11 -prepare" to start a manual Update, the command fails because the Version is not allowed on SV1.

Should I specify a Version 8.5.0. and still add the new nodes with version 8.7. ? 

BR
Matthias

Hello!

Yes, this can be done using what is referred to as Paced CCU or Manual Update.

It is a combination of these two procedures:
IBM Docs: Updating the System manually
and

IBM Docs: Replacing nodes nondisruptively wiht an IBM SAN Volume Controller SV3 node


What this does:

The Hardware replacement procedure prep steps readies your WWPNs and Port Number and makes sure the cluster pre-reqs are met

Then before you actually do the add node commands, you set the -pacedccu flag on the SV3 nodes, this will prevent the SV3 from downgrading it's agent code to match the cluster, it WILL downgrade it's cluster state machine to match the cluster.  This will allow you to run an 8.6.x.x version SV3 in a cluster at 8.5.0.x.

Repeat with every node in the cluster.

The tricky part is that you are reducing the number of IO Groups, so you will need to collapse down/remove the SV1 nodes entirely before the upgrade can be completed, and that may be difficult as you are not meant to go changing configuration during the upgrade process, which means you will need to migrate the IO Groups down first.  It is worth raising a support ticket to get walked through this, as they have been working through some pitfalls for other installs of this type.

I hope that helps.

Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

Hi All, 

One of our SVC clusters consists of 8x SV1 nodes and runs on 8.5.0.x
We are looking at replacing these with 4x SV3 nodes.
We want to do this via Node Replacement procedure.

We want to configure 3x 64Gb FC adapters in the SV3 nodes, but these are only supported as of 8.6.0.x, and the SV1 nodes cannot have 8.6.x.x.
Is there a way to still use the Node replacement procedure?
At the moment we don't have enough free fc-ports to create a complete new cluster.

Thanks.

------------------------------
TMasteen
------------------------------