HPE 9000 and HPE e3000 Servers
1753471 Members
4448 Online
108794 Solutions
New Discussion юеВ

Re: rp7410 cell electircal isolation question

 
SOLVED
Go to solution
D Bridges
Occasional Contributor

rp7410 cell electircal isolation question

Hello,

We are developing an HA Solution using 2 rp7410s and a va7410.

We will only have one nPartition per server.

We are contemplating purchasing a 2nd coreIO and then splitting our LAN card interfaces and FC cards between chassis 0 and 1 to provide more redundancy within each server, preventing unnecessary SG package switches.

My question:

With just 1 nPartition on a server, does the rp7410 continue to operate if 1 of the 2 cells/chassis fails?

From all the literature I have read, clearly if you have 2 nPartitions this would be the case. I cannot find any specifics regarding the behavior of a 1 nPartition, 2 cell system.

If you know of a specific document or thread on this, please let me know.

Thanks!
2 REPLIES 2
Ted Buis
Honored Contributor
Solution

Re: rp7410 cell electircal isolation question

If you have one nPartition with two cells and lose the entire cell the OS will crash.
There is some ability to lose memory chips (chip sparing) and sometimes dynamic CPU deallocation will save you from a crash, but consider if your I/O was off the failed cell, there would be no way to reach it from the other cell. So when there is not isolation between cells when they are formed into a single nPartition, only when they are in formed into separate nPartitions.
Mom 6
D Bridges
Occasional Contributor

Re: rp7410 cell electircal isolation question

Ted,

Thanks a bunch.

I have read the rp7410 white paper and have browsed most of the other docs associated with the rp7410, but I cannot find anything in writing.

I really need some documentation regarding the usage of 1 or 2 cells in a single nPartion system. If you know where to find this please let me know. I have looked through every doc on the rp7410 available on the docs.hp and am looking through the PartionsGuide this morning.

Regards,
Dewayne