WeeklyCeph

Ceph cluster health check

12 items ·5 critical ·3 warn ·4 info

Ceph health check

Weekly Ceph-specific check. Ceph is mostly self-healing, but proactive review catches slow-buildup issues (degraded PGs, failing OSDs, capacity drift).

Notes

HEALTH_WARN is acceptable if the cause is known and being addressed
HEALTH_ERR requires immediate attention
Latency trends are more useful than absolute values

Critical5 items

ceph -s reports HEALTH_OK
```
ceph -s
```
OSD tree shows all OSDs up and in
```
ceph osd tree
```
All PGs in active+clean state
```
ceph pg stat
```
All monitors in quorum
```
ceph mon stat
```
At least one manager active (HA mode: two)
```
ceph mgr stat
```

Warning3 items

Pool usage below 80% (warn at 70%)
```
ceph df
```
OSD commit and apply latency within expected range
```
ceph osd perf
```

SMART status of every OSD disk healthy

for osd in $(ceph osd ls); do host=$(ceph osd metadata $osd | grep "host": | awk -F\" "{print $4}"); ssh $host smartctl -H /dev/$(ceph osd metadata $osd | grep "osd objectstore" | awk -F\" "{print $4}"); done

Info4 items

Scrub and deep-scrub schedules running on time
```
ceph osd pool ls detail | grep scrub
```
Capacity trend reviewed for the past 30 days

OSDs evenly spread across hosts (no imbalance)

ceph osd tree --format json | jq -r ".nodes[] | select(.type==\"host\") | .children | length"

Ceph config backed up outside the cluster