Best scenario

[David_492]

Dear all


We have two HP UNIX Server rp8400. And we make the following steps:-

a. Each of these HP Server has two I/O Card each I/O Card has one SCSI Port and one Ethernet port plus the management port (local and remote). Also each server has one standby card for the network.

b. Each of these servers has 8 cpu 750mhz and 4 hard disk (36GB) and 16GB RAM.

c. We also bring two Disk (HP2300) arrays to act as external storage each array contains 4 Hard disks (36GB) EACH OF THESE arrays has 2 SCSI port.

d. We partition one of these two servers into two nodes (hardware partion) npar Node1 and Node2.

e. My question. What is the best scenario in this environment in order to achieve the clustering and no spf (single point of failure) between the three nodes (node1, node2, the un partitioned server).

[Michael Steele_2]

rp8400 SPOF are similar to a superdomes, so
system clock, power monitor, Core I/O backplane, and power backplane will all be vulnerable.

You receive better fault tolerance from stand alone servers then from partitioned servers, so SPOFs will exist in the nPar configuration between node1 and node2 but not the unpartitioned server, node3.

You have mentioned MC/ServiceGuard? If MC/SG then an externel cluster disk with an external power source is needed. This disk must stay up incase any node goes down.

[Stefan Farrelly]

For full redundancy (no SPF) you will have to have the rp8400 without npartitions as the backup/failover server for the rp8400 with x2 npars. This is the normal procedure for MC/SG clusters - always have completely separate hardware for failover/backup.