storage

SMR drives, ZFS and 92% full, and why compression won't save you

A ZFS mirror pool on shingled drives stalled under build load. The instinct is to reach for compression or a bigger cache, and both are useless here, because the stall was never bandwidth-bound. Measured at single-digit MB/s while completely wedged. Here's how to tell seek starvation from a throughput problem, and why the fix is moving the workload rather than tuning the pool.

zfsstorageinfrastructureperformance