Ceph health monitoring


Overview

XorMon-NG collects health data from the Ceph manager Prometheus module (default port 9283).

Area Behavior
Health Status page   Shows active Ceph health checks and errored cephadm daemons
Device status Driven by the default baseline checks below
(plus errored daemons), unless you customize Advanced settings

Official references:


Default behavior (no Advanced settings)

Area Behavior
UI All active health checks and errored daemons are visible on the Health Status page
Device status Default baseline checks (see table below) and any errored cephadm daemon
Ignored by default   CEPHADM_PAUSED is not counted toward device status

Default baseline

These check names must match the name label in Prometheus ceph_health_detail exactly.

Check name Typical meaning Ceph documentation
CEPHADM_FAILED_DAEMON   One or more cephadm daemons failed (for example osd in error state)   CEPHADM_FAILED_DAEMON
OSD_DOWN OSD is down OSD_DOWN
OSD_FULL OSD is full OSD_FULL
POOL_FULL Pool is full POOL_FULL
PG_DAMAGED Damaged placement groups PG_DAMAGED
MON_DOWN Monitor is down MON_DOWN
MDS_ALL_DOWN All MDS daemons are down MDS_ALL_DOWN
PG_RECOVERY_FULL Recovery blocked because cluster or OSD is full PG_RECOVERY_FULL

Typical cephadm failure: HEALTH_WARN with CEPHADM_FAILED_DAEMON and a daemon in error state is covered by the default baseline — no Advanced settings required.


Advanced settings (per Ceph device)

Configure in Settings → Device → Storage → your Ceph device → Advanced settings (key/value pairs stored per device).

Key Purpose Example value
health_extra_checks Add check names to the default baseline PG_NOT_SCRUBBED,PG_DEGRADED,POOL_NEAR_FULL
health_ignore_checks Remove check names from the effective set OSD_DOWN
health_alert_checks Replace the default baseline — only listed checks  
(plus errored daemons) count toward device status  		 CEPHADM_FAILED_DAEMON  

The key name health_alert_checks is fixed in the product; it means which checks affect device status, not a separate alerting UI on this page.

Value format

  • Use exact check names from Ceph (uppercase letters, digits, underscores), for example CEPHADM_FAILED_DAEMON.
  • Do not paste CLI messages (for example daemon osd.31 on node).
  • Separate multiple names with commas.
  • Discover names on your Ceph:
curl -s http://<mgr-host>:9283/metrics | grep 'ceph_health_detail{name='

Configuration examples

Default (recommended first test)

No Advanced settings. Baseline includes CEPHADM_FAILED_DAEMON and other critical checks.

Add monitoring for scrub backlog

Key Value
health_extra_checks   PG_NOT_SCRUBBED,PG_NOT_DEEP_SCRUBBED

Exclude OSD_DOWN from device status for this device

Key Value
health_ignore_checks   OSD_DOWN

Monitor only cephadm failures

Key Value
health_alert_checks   CEPHADM_FAILED_DAEMON

Suggested optional checks (not in baseline)

Often added via health_extra_checks when a site wants more sensitivity. See Ceph Health Checks for full descriptions.

Check name Note
PG_NOT_SCRUBBED   Long-running WARN on busy clusters; common false-positive candidate
PG_NOT_DEEP_SCRUBBED   Same as above for deep scrub
PG_DEGRADED Normal during recovery; may flap
SLOW_OPS Often transient under load
POOL_NEAR_FULL Capacity planning; usually WARN

References