I have searched via google but did not find any information on the read balancing algorithm for MD raid1.
Some years ago i have maybe read somewhere that it uses
hashing the process id to determine which device it uses, but i am not certain.
I have KVM virtualisation + LVM on top of raid1 and i witnessed multiple times that reads from multiple virtual machines go to the same disk, going 100% utilization and the other disk in the raid1 is idle. (via iostat -x)
EDIT: the question is also about multiple processes inside the VM. Will the read balancing happen even if inside the VM?
Are the reads balanced? If yes by which algorithm?
The reason i am asking is because i am currently dedicing between two variants
- LUKS on top of single MD device
- MD on top of two LUKS devices
The second variant uses more CPU ,but that is not a concern.