Application error monitoring

cron/servers_data_fetch.php records fatal-error and log-size events from its forced GET /apps response. Dashboard rendering displays those summaries.

Deploy the dashboard before the node. The dashboard accepts old node payloads during the rollout; legacy app flags are ignored and omitted monitoring means every check is enabled.

Node application settings

Configure optional per-app switches in mata-node config/apps.json. Every setting defaults to true; the node emits only false settings in its /apps response.

{
  "shop": {
    "monitoring": {
      "loganalysis": false,
      "logsize": false,
      "sessions": false,
      "reachability": false,
      "alarms": false,
      "events": false,
      "libraryaudit": false
    }
  }
}

• loganalysis, logsize, and sessions skip their respective node scans and omit status, logs, and session from /apps.
• reachability excludes the app from availability probes.
• events suppresses all app events.
• alarms records events but prevents their alarms.
• libraryaudit stops the scheduled Composer audit and its security events for the app. Manual audits from the Audit page still work.

An omitted result means not measured, not zero. The dashboard shows disabled checks as unavailable and removes log modal triggers. The node rejects the matching detail route with HTTP 409 when loganalysis, logsize, or sessions is disabled.

The dashboard server configuration flags error_monitoring_enabled and logsize_monitoring_enabled remain server-scoped. They combine with the app's loganalysis and logsize settings respectively.

Fatal events are occurrences. A positive fatal count has an opaque cursor; an event is recorded only when that cursor changes. A valid zero count clears stored cursor state and does not resolve, delete, or acknowledge historical alarms. Fatal events have no event cooldown by default; alarm delivery remains throttled for one day.

Log-size monitoring evaluates the largest /apps.logs entry against log_filesize_kb. It keeps the existing event and alarm cooldown behavior.

Library security advisories

cron/libraries_audit.php runs every 6 hours for apps with composer enabled on the node. It reads each app's Composer audit through the shared API cache, so the node runs composer audit only when the cached audit is older than the audit TTL in ttl.json (default 24 hours). Audits started from the Audit page refresh the same cache.

The job records library_outdated_security when an audit reports an advisory that was not in the previously evaluated audit for that app. The event is critical when a new advisory is critical or high, otherwise warning. Seen advisory IDs are stored in app_library_audit_state:

• Unchanged advisories record nothing.
• A fixed advisory leaves the stored set without an event; if it returns later, it is reported again.
• A clean audit clears the stored set.
• A failed or missing audit leaves the stored set untouched.