ZFS is explicit about trouble: zpool status will tell you a pool is DEGRADED and name the FAULTED device. Like a degraded RAID array, the pool keeps serving data, but you've lost your safety margin until it resilvers.
Read the status
zpool status -v tank # state: DEGRADED # sdb FAULTED too many errors # note the READ/WRITE/CKSUM error counts per device
Replace and resilver
zpool replace tank sdb sdd # swap faulted sdb for new sdd zpool status tank # "resilver in progress" # if the disk was only transiently flaky: zpool clear tank sdb # clear errors without replacing
Use zpool clear only when you're confident the errors were transient (a loose cable, a one-off bus reset) — a disk racking up CKSUM errors is failing and needs replacing. Run periodic zpool scrub so ZFS finds latent corruption before a disk death does.
Practise it
Pool degradation is one of the simulator's storage incidents, alongside RAID rebuilds. Work them in Linux troubleshooting practice.