Re: Cache battery and Raid Controller Design WAS: Re: Rép. : Question of a Newbie to POWER Systems

And since the batteries only power the Cache the question that needs asking is: "What happens if the power goes out and the system sits for 'too long'?" And the answer is: the battery dies, the cache data is lost, the ASP is lost, the system needs to be restored. IBM does not publish (at least clearly) the amount of time said battery will protect the cache but they do say: "Several Days." So it is important that you obtain replacement electrons for those no longer flowing into your system to allow the write caches to be flushed. This isn't normally a problem as most outages are short. However if you live in Tokyo, or New Orleans, or the entire North East of U.S. and South East Canada or other places that have experienced long outages you need a plan to get power to the system before those batteries go flat.

- Larry "DrFranken" Bolhuis

On 3/30/2011 12:02 PM, John Jones wrote:

Exactly. The cache battery is supposed to support the cache across power
loss events, not to flush the cache before the drives spin down. This is
especially necessary since the drives may be in a different frame (with
different power supply) than the RAID controller.

On Wed, Mar 30, 2011 at 9:56 AM, Chris Bipes<chris.bipes@xxxxxxxxxxxxxxx>wrote:

But if the Cache is kept live on the controller, when you power the system
back on, the controller will write the contents of the cache to disk, as the
disks are now powered up and spinning. This happens before you actually
start the OS, well the partitions at least. It is done is Hardware/Firmware
on the RAID controller, regardless of the OS being loaded.

--
Chris Bipes
Director of Information Services