The clean PG state indicates that all copies of a placement group's data are present and consistent across all acting OSDs. Together with active, it forms active+clean, which is the fully healthy state for a PG.

What clean Means Precisely

A PG is clean when:

1. The current number of replicas equals the desired pool replica count
2. All objects in the PG are fully replicated across all acting OSDs
3. No recovery or backfill is in progress for this PG
4. The acting set matches the up set (no remapping)

active+clean is the Goal

Every PG should eventually reach active+clean. When ceph status shows all PGs in this state, the cluster is fully healthy:

ceph status
# pgmap: 256 pgs: 256 active+clean; ...

Check the count:

ceph pg stat
# 256 pgs: 256 active+clean

# In JSON
ceph pg stat --format json | jq '.num_pgs_by_state[] | select(.state_name == "active+clean")'

clean vs Other States

Reaching clean After Degradation

After an OSD failure and recovery, PGs transition through several states before reaching clean:

active+degraded -> active+degraded+recovering -> active+clean

Watch the transition:

watch ceph pg stat

How Long Should It Take?

Recovery time depends on:

• Amount of data to recover
• OSD I/O capacity
• osd_recovery_max_active setting

Estimate using the bytes to recover:

ceph status | grep "degraded\|recovering"
# degraded (x%) objects, x/y bytes

Per-PG Clean Check

Inspect a specific PG:

ceph pg <pg-id> query | jq '{state: .state, last_clean: .info.history.last_epoch_clean}'

Pool-Level Clean Status

Check how many PGs per pool are clean:

ceph pg ls-by-pool mypool | awk '{print $1, $16}' | grep -v "active+clean" | head -20

Triggering Scrub to Verify Clean

Even a clean PG might have silent data corruption. Verify by triggering a deep scrub:

ceph osd pool deep-scrub mypool
ceph health detail | grep -i "inconsistent\|error"

Summary

The clean PG state confirms that all desired replicas are present and fully synchronized. The target healthy state for every PG is active+clean. When any PG deviates from this state, Ceph immediately begins recovery. Monitor how many PGs are in active+clean to track cluster health after OSD failures or maintenance events.