Network switch change from Juniper to Cisco

[MohitChordia]

Our network team is moving from Juniper to Cisco , they are going to change network switches to which our on-prem NetAPP array is connected which in turn will result in different IP addresses for data LIFs, intercluster LIFs , node and cluster management and Service processor.

a) Has anyone performed this type of change ?

b) How much downtime would this cause , seems like NFS would not be accessible by hosts and snapmirror relationship would break ?

c) I guess i need to open firewall ports for NFS and snapmirror again with new IP address.

c) Do we need to run snapmirror initial copy again, i hope it would take long time ?

d) What if i ask them to move entier VLAN and provide same IP on new switches, would it be non-disruptive ?

Regards,

Mohit

[aborzenkov]

D) it can be done non disruptively as long as you can do it port by port. Migrate LIFs off port, reconnect port to new switches, migrate LIFs back. Honestly, this is the only sane way to do it.

If switch migration is rip and replace, this is obviously downtime (unless your team is capable to perform replacement in less than a minute).

IP change means inevitable downtime for each client to switch to new IP.

You can resume SnapMirror without performing full sync. Also after IP change.

[MohitChordia]

Thanks for the reply,

Migrate LIFs port by port is certainly a nice way to do this, i will explore this option.

We have two different VLANs ( given by network team) . VLAN A has ip address assigend for NFS and snapmirror LIF whereas VLAN B has IP address assigned for node management, cluster management ( 2 node cluster) and service processor.

Iam sure migrating IP address of VLAN A ie NFS and snapmirror LIF would cause downtime .

can we migrate IP address of VLAN B without any downtime as they are management IPs only ?

Regards,

Mohit

[aborzenkov]

Only you know your infrastructure and only you can decide how disruptive it will be. You may be using management interfaces for NDMP or you may be having older Data ONTAP which was using node management LIFs for some control traffic associated with data access, so prolonged unavailability of LIF may impact data access as well, to name a few.

There is always route via takeover/giveback, in this case you recable node offline while keeping data access to other node.

[MohitChordia]

Thanks for the reply ,

We dont have NDMP or anyhting else connected to management interface.

As management VLAN is different from data and intercluster LIFs VLAN . Thus i assume that it they dont have any interdependency.  snapmirror and NFS LIFs should not be impcated while changing cluster management IP.

After discussing with network team , we found that VLAN A is isloated with netapp IPs ( DATA and INTERCLUSTER LIFs IPs) thus we would perform lift and shift migration ie migrating entire VLAN to new network. we hope this would cause minimum downtime as we are not changing IP address for data and intercluster LIFs. Do you think its correct ?

Regarding management IPs , they are with bunch of other hosts so we have to re-ip node, cluster and service processor network. Hope this would be online and would not cause any impact to NFS and Snapmirror operations. your comments ?

Regards,

Mohit