The undersized PG state indicates that the acting set for a PG has fewer OSDs than the pool's size setting, but the PG has NOT fallen below the min_size threshold. The PG continues to serve I/O but with reduced replication.

undersized vs degraded

These two states are closely related but have distinct meanings:

A PG can be undersized without being degraded if the acting set is smaller than size but all objects in the acting set are consistent. Conversely, a PG can be degraded without being undersized.

What Causes undersized

1. Not enough OSDs are available to fill the acting set
2. Insufficient OSDs in a CRUSH failure domain
3. Pool size is set higher than the available OSD count

# Check pool size settings
ceph osd pool get mypool size
ceph osd pool get mypool min_size

# Check how many OSDs are up
ceph osd stat

Checking Undersized PGs

ceph status
# HEALTH_WARN: X/Y pgs undersized

ceph pg stat | grep undersized

# Detailed list
ceph pg dump | grep undersized

For a specific PG:

ceph pg <pg-id> query | jq '{state: .state, acting: .acting, up: .up}'

Risk Assessment

Undersized PGs have less protection than configured. For a size 3 pool:

• 2 acting OSDs: can lose 1 more OSD before data loss
• 1 acting OSD: data is vulnerable

# Count undersized PGs by acting set size
ceph pg dump --format json | jq '.pg_stats[] | select(.state | contains("undersized")) | .acting | length' | sort | uniq -c

Resolving undersized

Add More OSDs

If the pool size exceeds available OSD count, add more OSDs:

ceph osd tree   # count available OSDs per host/rack

Reduce Pool Size

If the pool size is intentionally larger than available OSDs:

ceph osd pool set mypool size 2
ceph osd pool set mypool min_size 1

Recover Failed OSDs

If OSDs are down but recoverable:

systemctl start ceph-osd@<id>.service
watch ceph osd stat

Undersized During Maintenance

During rolling maintenance, undersized is expected and temporary. Set noout before starting:

ceph osd set noout
# Perform maintenance - PGs become undersized temporarily
# Restart OSDs
ceph osd unset noout
# PGs recover to full size

Summary

Undersized PGs have a smaller acting set than the configured pool size, which means reduced but not necessarily absent redundancy. The cluster continues serving I/O but with less fault tolerance. Resolve by adding OSDs, recovering downed ones, or adjusting the pool size to match the available hardware. During planned maintenance, this state is expected and transient.