Skip to content

Agent self-logs

Self-logs bring the agent’s own error output into KloudMate, so you can diagnose a failing agent without opening a shell on the host. Like the health metrics, they are shipped out-of-band, on a path independent of the collector. So an error survives even when the collector cannot run. Self-logs apply on all deployment modes.

The agent forwards its own ERROR-level logs directly to KloudMate. It ships only errors, not routine debug or info messages. Each log record carries attributes that tell you where the error came from:

  • The source within the agent, such as the check-in, discovery, instrumentation, or the collector.
  • The host or node the agent runs on.
  • The agent version.

On Kubernetes, self-logs are shipped per node, each tagged with its node identity. So you can see errors from each node’s collector and from the fleet manager without running kubectl logs. See Kubernetes platform notes.

Self-logs are on by default, controlled by the KM_AGENT_ERROR_LOGS setting.