Cluster Configuration

COG · ‎2018-04-25

Hi All,

I am relatively new managing Netapp storage systems. Any help suggestions I can help will be much appreciated.

We currently have 10 controllers and a network configuration that has all the lifs on a particular vlan in the same broadcast domain and failover group. However we are about to replace 4 fas8080 filers with 4 fas9000 filers. The issue is that our network team is also moving into a new network configuration whereby the old vlans will no more be used in the new configuration and a network folk said that the new vlans on the new core will be layer 3 vlans while the old ones are layer 2, but data going to the old vlans can be routed to the new core switch. We plan to connect these 4 new filers to a new core switch in the new configuration and this core will not have any of the old vlans configured on it. There seem to be two options (as we will prefer not to have the new filers connected to the old network since we will soon migrate everything over to the new network configuration):

1. Follow this route above and join the new nodes to the same cluster as the old ones but have broadcast domains and failover groups of the new nodes that are different from the those of the others in the old vlans. So failovers will be between 4 nodes instead of 10.

2. Avoid the use of vlans all together in configuring the lifs and configure lifs on top of ifgrps. Is this possible if the switch ports are configured as access ports instead of trunked ports? What about traffic to/from the intended clients for the different vlans getting/sending their data?

Please does anybody envisage any problems down the line, especially with vservers, data traffic, snapmirror, snapvaulting, etc.?

Examples

broadcast domains
Default 136.171.74.0/24 1500
                            node1:a0a-1000             complete
                            node2:a0a-1000             complete
                            node3:a0a-1000             complete
                            node4:a0a-1000             complete
                            node5:a0a-1000             complete
                            node6:a0a-1000             complete
                            node7:a0a-1000             complete
                            node8:a0a-1000             complete
                            node9:a0a-1000             complete
                            node10:a0a-1000             complete
        146.36.200.0/21 1500
                            node7:a0b-2000              complete
                            node3:a0b-2000              complete
                            node2:a0b-2000              complete
                            node6:a0b-2000              complete
                            node4:a0b-2000              complete
                            node2:a0b-2000              complete
                            node1:a0b-2000              complete
                            node5:a0b-2000              complete

Failover Groups
node00
                 136.171.74.0/24
                                  node1:a0a-1000, node6:a0a-1000,
                                  node2:a0a-1000, node7:a0a-1000,
                                  node3:a0a-1000, node8:a0a-1000,
                                  node4:a0a-1000, node9:a0a-1000,
                                  node5:a0a-1000, node10:a0a-1000
                 146.36.200.0/21
                                  node1:a0b-200, node5:a0b-200,
                                  node2:a0b-200, node6:a0b-200,
                                  node3:a0b-200, node7:a0b-200,
                                  node4:a0b-200, node8:a0b-200

Thanks.

aborzenkov · ‎2018-04-25

@COG wrote:

I am relatively new managing Netapp storage systems.

This question has very little to do with NetApp, it is mostly networking question The only thing you need to ask your networking team - whether new ports will be in the same logical IP network or not. If they are in the same network - they should be added to the same broadcast domain on NetApp side. If they are in different IP network - this implies full scale migration, as it affects much more than just broadcast domain configuration on NetApp.

@COG wrote:

new vlans on the new core will be layer 3 vlans while the old ones are layer 2

It is absolutely unclear what it means. There is no such thing as L3 VLAN. It may mean routing between different VLANs over L3 (IP) interfaces or it may mean bridging different VLANs over L3 nertwork (VXLAN as example).

@COG wrote:

1. Follow this route above and join the new nodes to the same cluster as the old ones but have broadcast domains and failover groups of the new nodes that are different from the those of the others in the old vlans. So failovers will be between 4 nodes instead of 10.

This implies different IP network on new nodes.

@COG wrote:

2. Avoid the use of vlans all together in configuring the lifs and configure lifs on top of ifgrps.

You probably misunderstand how it works. Every bit that is carried by switches belongs to some VLAN. Whether this VLAN is exposed to connected host (tagged VLAN, VLAN interface on NetApp) or implicitly associated with physical port (untagged VLAN, physical port/interface group on NetApp) is irrelevant. Of course you as NetApp admin must know how switch is configured to match NetApp side, but it has nothing to do with original question - whether interface belongs to the same IP network (and hence logical broadcast domain) or not. In case of VXLAN local VLAN numbers on two switches may be different, but both ports still will belong to one and the same broadcast domain.

COG · ‎2018-04-27

Thanks for your response. This is the definition of a broadcast domain - Broadcast domains enable you to group network ports that belong to the same layer 2 network. Please could you elaborate on the logical IP part viz-a-viz this definition of a broadcast domain? Are you referring to subnets and/or vlans?

Also could you elaborate on this - "If they are in different IP network - this implies full scale migration, as it affects much more than just broadcast domain configuration on NetApp". This is one of the questions I posted.

aborzenkov · ‎2018-04-28

@COG wrote:

Broadcast domains enable you to group network ports that belong to the same layer 2 network.

What I tried to tell - Layer 2 network is not always synonym for VLAN. Today Layer 2 network may well span multiple switches with different VLANs each.

@COG wrote:

Also could you elaborate on this - "If they are in different IP network - this implies full scale migration, as it affects much more than just broadcast domain configuration on NetApp".

You will need to add or change IP addresses of your SVMs, so whatever needs to be done to make clients use new addresses.

COG · ‎2018-04-30

Thanks and I appreciate your attempting to help. It would be ore helpful for me though if you or anybody else could be more elaborate or provide examples as this is the first time I am going to do this.