The degraded PG state means one or more object replicas are missing. The PG is still serving I/O (when combined with active) but has reduced redundancy. Understanding this state is critical for assessing data durability during and after OSD failures.

What degraded Means

A PG is degraded when its current replica count is below the pool's size setting. For a size 3 pool, degraded means the PG has 2 or fewer copies.

Degraded does NOT mean data loss - it means you have less protection than configured. If another OSD fails while a PG is degraded, you could lose data if the remaining replicas go down.

Checking Degraded PGs

ceph status
# HEALTH_WARN: X/Y objects degraded

ceph pg stat
# active+degraded count

# List which PGs are degraded
ceph pg dump | grep degraded

Get the number of degraded objects:

ceph status --format json | jq '.pgmap.degraded_objects'

What Causes Degradation

1. An OSD goes down or out
2. A node is rebooted during maintenance
3. An OSD is being replaced
4. Network partition preventing replica communication

Check which OSDs are affected:

ceph osd tree | grep -E "down|out"
ceph health detail | grep "pg"

Degradation Levels

Risk Assessment

With one replica missing (2/3), the cluster can tolerate zero additional OSD failures for affected PGs. With two replicas missing (1/3), any single additional OSD failure could make those PGs inactive.

# See degraded ratio
ceph status | grep "degraded"
# degraded (33.333%): 1024/3072 objects

# Calculate affected PGs
ceph pg dump | awk '{if ($16 ~ /degraded/) print $1}' | wc -l

Speeding Up Recovery

If degradation is due to a recoverable OSD coming back:

# Check if OSD is recovering
ceph osd stat

# Increase recovery speed
ceph config set osd osd_recovery_max_active_hdd 10
ceph config set osd osd_recovery_op_priority 10

If the OSD is permanently lost, mark it out to trigger remapping:

ceph osd out osd.3

Monitoring Recovery Progress

watch ceph status
# Look for transition from:
# active+degraded -> active+recovering+degraded -> active+clean

Summary

The degraded PG state indicates reduced data redundancy due to missing replicas. Degraded PGs continue serving I/O but with lower fault tolerance. When you see active+degraded, act promptly by identifying the failed OSD, determining if it is recoverable or must be replaced, and ensuring the remaining replicas do not fail during recovery.