Health checks monitor device collector resources on a self-managed Director and are configured in vmetric.yml. The timing constants below are not configurable at all; they are documented for operational awareness.
Health Check Configuration
Health checks monitor device collector resources. A resource must exceed its threshold for the configured number of consecutive iterations before action is taken.
Check Toggles
| Field | YAML Path | Type | Default | Description |
|---|
| Check CPU Usage | health.check_cpu_usage | bool | false | Monitor CPU usage |
| Check Memory Usage | health.check_memory_usage | bool | false | Monitor memory usage |
| Check Log File Usage | health.check_log_file_usage | bool | true | Monitor log file sizes |
| Check TMP Usage | health.check_tmp_usage | bool | false | Monitor temp directory |
| Check Report Usage | health.check_report_usage | bool | true | Monitor report files |
| Check Simulator Usage | health.check_simulator_usage | bool | true | Monitor simulator |
| Check Hanged Processes | health.check_hanged_processes | bool | true | Detect hung processes |
| Check Stats Usage | health.check_stats_usage | bool | true | Monitor stats collection |
Thresholds
| Field | Default | Description |
|---|
| CPU Limit | 80% | CPU usage threshold percentage |
| Memory Limit | 1024 MB | Memory usage threshold |
| Max Log Size | 10 MB | Maximum log file size |
| Stats Retention | 24 hours | Stats retention period |
Threshold Iterations
All threshold iteration defaults are 10. A resource must exceed its threshold for this many consecutive check iterations before the health system takes action.
Internal Timing Constants
These values are hardcoded and not user-configurable. They are documented here for operational awareness.
Manager Intervals
| Constant | Value | Purpose |
|---|
| Heartbeat Update Interval | 5 s | Device heartbeat publishing frequency |
| Leader Check Interval | 1 s | Leadership election check frequency |
| KV Sync Interval | 5 s | JetStream KV synchronization |
| Config Sync Interval | 1 s | Configuration change polling |
| Heartbeat Check Interval | 15 s | Device health check and node offline threshold |
Cluster Update Intervals
| Constant | Value | Purpose |
|---|
| Cluster Update Check Interval | 10 s | Poll for pending updates |
| Node Timeout | 30 min | Maximum wait for a single node update |
| Health Check Wait | 30 s | Post-update health check delay |
| Leader Stability | 30 s | Required leader stability before proceeding |
| Command Retry Interval | 30 s | Between retry attempts |
| Max Retries | 3 | Maximum update retries per node. The offline-node retry path reuses the same limit — there is no separate offline budget |
Upload Starvation
These govern when a target is reported for upload starvation. See Target Uploads Starving.
| Constant | Value | Purpose |
|---|
| Window | 10 min | Trailing window every condition is measured over |
| Sample Interval | 30 s | How often the detector samples |
| Minimum Flushes | 3 | Flushes a window must hold before its average fill counts |
| Fill Threshold | 5% | Average fill of batch_size under which flushes count as small |
| Queue Rise | 10 notifications | Least growth of the target's own queue across the window |
| Store Rise | 1 MiB | Least growth of the payload store across the window |
| Repeat and Clear | 10 min | The warning repeats this often while it lasts, and clears after this long without it |
Configuration Limits
| Constant | Value | Purpose |
|---|
| Config File Max Size | 1 MB | Maximum configuration file size |
| Retention Interval | ~3 months | Default data retention period |