Skip to content

System Health Checks

System Health Checks exposes the health state of Emakin components in JSON form. Use it to confirm system availability, detect degraded dependencies, and identify components that need attention.

The health endpoint is available at:

1
http://<servername>:5121

Result States

Result Description
Healthy Component is functioning correctly
Degraded Component has a non-critical issue
Unhealthy Component has failed

Cluster Report

The health endpoint reports the whole cluster, whichever node you query. It includes nodes running web, SMTP, or full-text services without workers. The same information is available in Host Administration's Worker monitoring tab.

The JSON response uses schema version 2:

  • status is the overall system health.
  • coverage indicates whether the report has current information for all visible nodes.
  • results lists shared application and service checks.
  • nodes lists individual processes, their roles, availability, measurement times, and health results.

Monitoring integrations that previously read CPU, thread count, or disk checks from results must now read nodes[].results.

Example response (abbreviated):

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
{
    "version": 2,
    "status": "degraded",
    "statusIcon": "⚠️",
    "coverage": "complete",
    "observedAt": "2026-11-12T10:00:00Z",
    "results": [],
    "nodes": [
        {
            "node": "web-1",
            "processId": "e45da16e-7fde-448d-8bfe-1bd7957eaf95",
            "roles": ["web"],
            "seenAt": "2026-11-12T09:59:58Z",
            "observedAt": "2026-11-12T09:59:55Z",
            "availability": "current",
            "metricsCurrent": true,
            "status": "degraded",
            "results": [
                {
                    "name": "Thread Count",
                    "status": "degraded",
                    "data": { "threadCount": 250 }
                }
            ]
        }
    ]
}

CPU usage is the process's CPU usage normalized by logical processor count, not total machine CPU utilization. Disk available and required values are in MB and describe the drive containing the application. Thread count belongs to the process. Processes on the same machine are listed separately, and process identity changes after a restart.

A failed connection on one node can affect the overall result even when other nodes can reach the same service. Inspect that node's results to identify the affected connection. Full-text catalog problems are shown under the node hosting the catalog.

Full-Text Availability

One full-text node is active; the others remain ready to take over. A passive role is normal and does not by itself indicate a problem.

Situation Overall full-text health Administrator action
Active service is available and standby nodes are ready Healthy No action required.
Active service is available, but a standby node is not ready Degraded Inspect the affected node's registration, storage access, and catalog checks.
No usable active service is available Unhealthy Check full-text service availability and restore an active service.

After failover, confirm that the replacement node is active and review the readiness of the remaining standby nodes.

The Cache check reports Redis connection availability for each node. It does not measure cache hit rate, memory usage, or operation latency.

Freshness and Availability

Information normally updates approximately every 10 seconds. A node shows delayed after 30 seconds without an update and unreachable after 60 seconds. Measurements older than 30 seconds are marked out of date. Check both node availability and measurement time: a visible node can still have missing or old measurements.

seenAt shows the last node update; each node's observedAt shows when its health measurements began. Missing or stale measurements prevent a healthy overall result. For worker nodes, this also applies when worker status is missing, out of date, or not running. Planned stopped processes remain visible but do not degrade the remaining running processes.

Coverage Meaning Administrator action
complete All visible running nodes have current information. Health checks may still be degraded or unhealthy. Review status and any failing checks.
partial Information for one or more nodes is missing or out of date. Check the affected service or pod and its connectivity.
unavailable A system report cannot be obtained, or no running nodes are visible. Check service availability and monitoring connectivity.

Healthy and degraded reports return HTTP 200; unhealthy reports return HTTP 503. Configure monitoring to inspect the JSON status and coverage as well as the HTTP status code, so degraded or incomplete reports are not missed.

Old processes disappear from the list after 10 minutes. An absent node is not proof that the expected deployment is healthy: compare the list with the services and replicas that should be running. Keep deployment-level availability alerts in place. During an upgrade, update every node; older versions may appear without health measurements.

Use service-specific checks for automatic restarts and traffic routing. A cluster-wide problem should not cause a healthy node to be restarted or removed from traffic.

The health endpoint allows anonymous access and includes operational details. Restrict access to trusted monitoring networks.

Documented Health Checks

The documented health checks include:

  • Disk Space
  • CPU Usage
  • Thread Count
  • Cache
  • Application Database Connection
  • Market Connection
  • Mobile Push Connection
  • Timestamp Service Connection
  • Full Text Search Service
  • Full Text Service Database Access
  • Full Text Service Registration
  • Full Text Catalogs
  • Mail Sending Jobs
  • Mail Receive Jobs
  • Workflow Processing Jobs
  • OpenAuth <servicename> Service Connection
  • FileRepository <servicename> Directory
  • Ldap <ldaphost> Service Connection
  • WOPI <servicename> Service Connection
  • Message Service <servicehost> Connection
  • Tika Service Connection

Use the detailed per-check configuration and thresholds in your environment as the canonical reference for exact failure conditions and severity levels.