Hello,
We have several clusters with spinning disks (with and without flash pool). On most of them we observe events for "node disk fragmentation" (high node utilization caused by high percentage of delayed disk write) despite that free space realloc is enabled everywhere. I have even tried to schedule from node shell the old-style full aggregate reallocation which should help for free space fragmentation but that's also not helping much (or for a long time).
How can I verify that the free space realloc is working and doing it's job ?