6 drives: RAID1 for /, RAID1 with write back cache enabled, RAID1 with write back cache enabled
A vendor came back to us and said that the load spikes we were seeing on a server were due to the RAID1 for the OS disks on / did not have write back cache enabled but the other two did. When they enabled it the load spikes went down and they stated that having it setup this way will "cause load calculation issues on Linux"
Has anyone heard this before? Their fix worked (enabled write back cache on all RAID devices through MegaCLI) but it seems counter intuitive to me that you cant have a mix of write back cache among multiple RAID devices on a single server - nor have I ever heard that before.
EDIT: For some more context, just about nothing is done on / so with BBU + write back cache for / being enabled would not have been such a drastic performance enhancement if you know what I'm saying
[link][11 comments]