14 node network setup redundancy/HA Best practices?

[KasperNico]

Hi,

We have a 14 node P4500 cluster, all in one MG, all in one CL and need more redundancy and HA.

We have a few things we would like to clear. Please see attached image.

- There is a HP Procurve mesh setup between Site A and Site B. Switches are of type 3500yl.

- All storage nodes are configured in Bond ALB/Flow control(also on switches)/1000Mbit Full.

- At the moment there is no FOM, so Site A is running 3 managers, while Site B is running 2 managers, at the moment Site B is allowed to go down, in the future the MG needs to maintain quorum at all time when site failure occurs.

- We dont want to use routers to route the traffic between 10.128.0.0 and 10.128.16.0 because it will produce more latency. We would like to use layer 2 routing instead.

- How do we define a default gateway to the nodes, because only one switch can be the gateway. Or do we need to use RIP fort his part? This in case a switch failure occurs.

- Will the setup with the FOM work over the IPSEC connection, all it needs to do is maintain quorum, right? There shouldnt be a problem if the FOM is gone for a short period of time, right?

- What will happen if the switch mesh(2 seperate fibers combined, marked A in the image) goes down? Will there be a split brain situation?

Advice would be higly appreciated!

Regards,

Nico

[Emilo]

- We dont want to use routers to route the traffic between 10.128.0.0 and 10.128.16.0 because it will produce more latency. We would like to use layer 2 routing instead.

Am I missing something? Dont we need a router to 'route' between vlans?

it looks like what you want to setup i s  a multi-stie cluster.

I cant explain everything here. So look at this manual and pay particular attention to the required bandwidth needed round trip latency needs to be no more then 2 ms between sites.

Look at page 5 under the section best practices.

Anyway read this document and pick a model then let me know any question s you might have.

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c03562494/c03562494.pdf

[oikjn]

what do you mean "more redundancy"? what are you trying to increase for redundancy? More copies of your data on more nodes or redundancy in case the management group goes down? It looks like you have the site setup correct (assuming you have the correct bandwidth requirements there).

I believe, but haven't tested, that your IPSEC FOM is fine. If that connection is lost for a little you won't have a split brain problem UNLESS you lose connection to the FOM and both sites also lose connection to eachother. Actual bandwidth usage for the FOM is tiny, but I think the manual says you need 100mbit, but from what I've seen, 10mbit is plenty.