rp7410 and internal redundancy

[Arnulf Gercke]

Hi,

customer runs an rp7410 with 4 CPUs and 4x2GB RAM on ONE!! Cellboard.
Because of 1 Cellboard he has only 1 I/O Cage containing 2x FC-HBA and additional NICs.
There are 2 standart PCI Power Supplies in the Server.
Question:
What happens, if the one PCI powersupply utilizing the I/O Cage fails?
Will the 2nd PCI power supply take over or will I/O stop working?
If it would stop working, will a 2nd PCI Cage and Cellboard with evenly distributed CPU/RAM/HBA solve the problem?

[Jeff Schussele · Accepted solution]

Hi Arnulf,

It's been my understanding that in the 7410 line the PCI power supply is inherently a SPOF - but only to I/O. The loss would only affect the I/O cage - not RAM or CPUs.
Therefore when we use 7410s in mission critical servers we double up our I/O across the cages & depend upon MC/SG (or APA) to failover the LAN card(s). And we depend upon LVM to switch disk I/O to the alternate link HBAs.
So in your case you have a definitive SPOF - until you install another cell board.
And BTW you *could* get a "blank" cell board & evenly distribute the existing CPUs and RAM across them. But *remember* you'd *still* need to double up your I/O as well.
Moral: True HA is never cheap.

My 2 cents,
Jeff

[Joshua Scott]

the rp7410 is not really meant to be a HA system, especially with only 1 Cell board.

The SPOFs with only one cell are:
Cell board
CC Fan (on the cell board)
System Backplane
MP card
Core I/O card
I/O Chassis
I/O Power Supply

By adding a second cell board and distributing the cards across both I/O chassis, the only SPOF would be the system board and MP card. The MP card is not hot-swappable in the rp7410, and doesn't fail-over to the second card.

Josh

[Arnulf Gercke]

Thanks!