Four node P4000 cluster stretched over two sites

[Patrick Terlisten]

Hello,

a customer wants to use standard instead of multi-site, so there will be a single VLAN/ IP-subnet over both sited. With two nodes no problem, but what about four nodes and two-way replication?

Thanks for you help.

Regards,
Patrick

[Steven Clementi]

As long as the bandwidth and latency requirements are met, it would probably work... but I think there would be no guarantee that mirrored data would go to one of the 2 nodes in the other site.


Steven

[Patrick Terlisten]

Hello Steven,

that's the point. As far as I remember correctly this is only guaranteed if you're using a multi-site cluster. But in this case, customer has to route the iscsi traffic between the VLANs and you have to use "fixed" path policy for VMware vSphere, so you can't use multipathing with the round-robin policy.

Regards,
Patrick

[Uwe Zessin]

You can set up a Multi-Site cluster with a single VIP.

[Patrick Terlisten]

Hello Uwe,

I thought that two VIPs are mandatory for a multi-site cluster, as well as two sites and two subnets?! Is there any doc where this is described?

Thank.

Regard,
Patrick

[Uwe Zessin · Accepted solution]

P4000 Multi-Site HA/DR Solution Pack userguide
AX696-96004

On page 1:
1 Designing a Multi-Site SAN
...
* Multi-Site clusters that @@can@@:
-- Span up to three sites
-- Span multiple subnets
-- Have multiple VIPs

On page 22:
* Creating Multi-Site clusters and volumes
...
Add one or more VIPs, depending on the number of subnets you have configured.


And I have set up a Multi-Site cluster with one VIP.

[Patrick Terlisten]

Hello Uwe,

thanks for clarification.

Regards,
Patrick

[kghammond]

We are doing this... Although our WAN is 10 GB fiber so I am not sure if you call it a WAN.

Anyway, we are using one VIP and one subnet. Technically we have Multi-Site setup from the LeftHand Management Console as we have defined multiple sites so the cluster reports as multi-site, but from talking to LeftHand, this is more logical than functional.

The real important part is understanding how two-way replication works with more than two nodes. Simply put, you need to make sure that the nodes listed in the "Edit Cluster" window are ordered Site A, Site B, Site A, Site B.

This will have the effect that each two way replication node will have site resiliency, but it does not mean you can lose one node from each site without LUN's going offline. You can lose one node or both nodes from one site and keep all your LUN's online.

I hope this makes sense.

Kevin