Im about to monitor my storage performance, so I have some very rough idea whether my system performance is degrading or not (worn or too full / fragmented drive).
My storage setup is 2x 4TB NVME SSD in ZFS RAID1 with VM and container volumes on it.
My idea is to run a fio script in unprivileged container like every week (in off hours to avoid peaks in disk access) and log the results and check them occasionally or when experiencing performance issues.
Any suggestion if this will do the job or how can I improve the script parameters (maybe to be more reproducible, reliable)? Again the main goal is relative comparison in time, not absolute benchmarking, but a very basic (general) performance comparison across nodes would be a nice benefit of course.
Thanks!
My storage setup is 2x 4TB NVME SSD in ZFS RAID1 with VM and container volumes on it.
My idea is to run a fio script in unprivileged container like every week (in off hours to avoid peaks in disk access) and log the results and check them occasionally or when experiencing performance issues.
fio --name=mytest \ --filename=/opt/zfs-test/testfile \ --size=1G \ --bs=64k \ --rw=randrw \ --direct=1 \ --runtime=30s \ --time_based \ --group_reporting --output=/var/log/fio-run.logAny suggestion if this will do the job or how can I improve the script parameters (maybe to be more reproducible, reliable)? Again the main goal is relative comparison in time, not absolute benchmarking, but a very basic (general) performance comparison across nodes would be a nice benefit of course.
Thanks!