As part of my performance testing of our new hardware I would like to build a 2 drive mirror and simulate a nearly full drive resliver to determine the time it would take to repair.
We'll be using ZFS mirrors to house all of our VMWare data but I am assuming on a striped mirror that wouldn't really matter that much (correct me if I am wrong).
I'm figuring a 2TB mirror would take way too long using /dev/urandom and I don't think using /dev/zero would help to find a worst case resliver.
Would using shred generate enough psuedorandom data to fill the drive? This will be run on OpenIndana for testing since we're doing an all-in-one with ESX and Napp-It.
I want to benchmark as much as I can while the system is under component testing as I can and this along with some bonnie++ numbers will give us something to go on moving forward and array rebuild time is a huge concern for me.
We'll be using ZFS mirrors to house all of our VMWare data but I am assuming on a striped mirror that wouldn't really matter that much (correct me if I am wrong).
I'm figuring a 2TB mirror would take way too long using /dev/urandom and I don't think using /dev/zero would help to find a worst case resliver.
Would using shred generate enough psuedorandom data to fill the drive? This will be run on OpenIndana for testing since we're doing an all-in-one with ESX and Napp-It.
I want to benchmark as much as I can while the system is under component testing as I can and this along with some bonnie++ numbers will give us something to go on moving forward and array rebuild time is a huge concern for me.