Create a Post
cancel
Showing results for 
Search instead for 
Did you mean: 
Highlighted

Migrating cluster from old to new hardware

Jump to solution

Hi,

We are finally replacing our FW cluster with old UTM appliances for 5600 appliances. I would like to keep the same names in the policy, but since the interface names change I would like to know what the best way is to migrate to the new appliances with minimal outage.

I was about to failover to HA -

- move cables from the Primary appliance to the new 5600 Primary appliance,

- migrate export of the policy. Then remove all references of the existing cluster from the policy and delete the whole cluster from the management server.

- create a new cluster with initially 1 member (the new primary 5600) establish SIC and configure cluster with all new interfaces - Add cluster to the rules where the old cluster was removed

- Remove cables from Old HA Firewall,while installing the policy to the new Primary

- connect new 5600 HA and add to the cluster (and install policy)

Any other (or better) recommendations for a smooth migration to the new hardware?

Or can I just delete 1 cluster member and add the new hardware with different interface names to the cluster object?

Many thanks.

0 Kudos
1 Solution

Accepted Solutions
Highlighted

It's then really straight forward:

1. Start with standby node and migrate configuration of the node to the new node:

      I would do this by using actual clish config, editing it to change interface names (best with search/replace)

      Don't forget to do other config changes not represented in clish to the new node (scripts/fwkern.conf/etc)

2. Disconnect old standby from network

3. Conncet new standby to network

4. Reset sic through smartconsole

5. Change version/appliance type in general properties of cluster object

6. Change advanced configuration of interfaces in the topology/network management part to allow different interface names on both nodes. It doesn't matter if the cluster interface name is the new one or the old one during migration. But I prefer to have the new one already.

7. Install policy

8. Failover to new node with new version using cphastop on active node

9. Repeat above steps 1-7 to 2nd old node.

View solution in original post

9 Replies
Highlighted

Is this a Full-HA Cluster (Management on same box as gateway)? Or is it a distributed environment (Management running on separate server)?

I am not sure as you are speaking of migrate export in your description.

If it is distributed, you can easily change the existing cluster object to have interface names adjusted to new names and also adjust appliance hardware type.

0 Kudos
Highlighted

Hi Norbert,

It's a distributed environment. Only hardware replacement of the cluster, so should be straight forward. Management is running R80.10 existing cluster R77.20 and new hardware will be running R80.10.

It's only that we need to keep the outage window as small as possible.

My only concern is the change in interface names and how to make that as smooth as possible.

The migrate export is basically only a safety measure to have a good configuration of the policies to rollback to if the migration fails.

So short: Is it easier to remove the whole cluster object from the policy and create a new one with the new hardware/interface names or can I remove one Firewall object from the cluster and add a new one in with the new interfaces, although both cluster members than have different interface names.

Thanks again.

0 Kudos
Highlighted

It's then really straight forward:

1. Start with standby node and migrate configuration of the node to the new node:

      I would do this by using actual clish config, editing it to change interface names (best with search/replace)

      Don't forget to do other config changes not represented in clish to the new node (scripts/fwkern.conf/etc)

2. Disconnect old standby from network

3. Conncet new standby to network

4. Reset sic through smartconsole

5. Change version/appliance type in general properties of cluster object

6. Change advanced configuration of interfaces in the topology/network management part to allow different interface names on both nodes. It doesn't matter if the cluster interface name is the new one or the old one during migration. But I prefer to have the new one already.

7. Install policy

8. Failover to new node with new version using cphastop on active node

9. Repeat above steps 1-7 to 2nd old node.

View solution in original post

Highlighted

Thanks Norbert,

I'll give it a go.

I have done similar replacement s in the past from open server to appliance or Solaris to appliance. It was always straight forward, but somehow this one (maybe because it looks different in R80.10) it confused me.

Thanks a lot.

Jan

0 Kudos
Highlighted

Any other considerations if using VRRP?

Thank You

0 Kudos
Admin
Admin

It should be fairly similar steps.

0 Kudos
Highlighted
Silver

I'll piggyback on the back of this post with another question.  I know mis-matched hardware in a cluster is not supported, but that aside, would it actually still work?

What is it about the cluster members that must be the same for ClusterXl to be happy?  I suspect # of CPU cores has to match?  What about RAM amount?  Interface names - I.e. could one server have eth0, eth1, eth2, etc. and the other server have WAN, LAN1, LAN2, etc?  

I'm wondering on the best way migrate from open server to appliance.  The hardware will clearly be different.  Can I add the one of the new appliances in to the existing Open Server cluster and flip over to it to cut out downtime?  Or are they so mis-matched that my only option is downtime and physical cable swap from old to new?

0 Kudos
Highlighted
Platinum

RAM is not taken into consideration in case ClusterXL. You can have one machine with 16 GB and other cluster member with 32 GB and you can have active-standby cluster.

Interface names are not taken into consideration in case ClusterXL. Only IPs are important. During policy push, the gateway will assign IPs to the actual interface on the machine despite you have configured different interface name in Topology.

Only CPU Cores are important, the machine with more IPv4 cores will became Ready. It might be also not relevant for R80.40+.

Kind regards,
Jozko Mrkvicka
Highlighted
Hello Jozko,

I am doing similar migration from 13500 appliances to 15400 models and need limited downtime. I am also thinking about less rollback complexity as well.

The 13500 and 15400 seem to display the same core count in cpview. I am unsure however if both would sync anyway so it would probably be best to disconnect the active member just before policy install.

Any thoughts on this?

The second method I am thinking about is to:
1. Disconnect the standby member early in the maintenance window and create a new cluster object with same cluster IP
2. Add new standby member to new cluster, get interfaces and complete topology setup and cluster settings.
3. Do a Where used and Replace to replace the old cluster object with the new one, set policy target to the new firewall cluster,
4. Disconnect the old active member and finally push policy on the new firewall
Some gotchas in my environment would be that we use identity agent so users will need to trust the new certificate of the gateway I believe.

Thanks for your input.
0 Kudos