I can only comment I the log. I would recommend using three logs (6 disks
as mirror pairs) per system, and add a crush map hierarchy level for cache
drive so that any given pg will never mirror twice to the same log. That'll
also reduce your failure domain.
On Jan 29, 2014 4:26 PM, "Geraint Jones"
Hi Guys
We have the current config :
2 x Storage servers, 128gb RAM, dual E5-2609, LSI MegaRAID SAS 9271-4i, each
server has 24 x 3tb Disks these were originally set as 8 groups of 3 Disk
RAID0 (we are slowly moving to one OSD one disk) We initially had the
journals stored on an SSD, however af
