Log Record Fields

The following fields can be present in log records exported by the logging. Although log records are typically formatted as JSON objects, the same data model can be applied to other encodings.

To search these fields from Elasticsearch and Kibana, use the full dotted field name when searching. For example, with an Elasticsearch /_search URL, to look for a Kubernetes pod name, use /_search/q=kubernetes.pod_name:name-of-my-pod.

The top level fields may be present in every record.

The original log entry text, UTF-8 encoded. This field may be absent or empty if a non-empty structured field is present. See the description of structured for more.

Original log entry as a structured object. This field may be present if the forwarder was configured to parse structured JSON logs. If the original log entry was a valid structured log, this field will contain an equivalent JSON structure. Otherwise this field will be empty or absent, and the message field will contain the original log message. The structured field can have any subfields that are included in the log message, there are no restrictions defined here.

A UTC value that marks when the log payload was created or, if the creation time is not known, when the log payload was first collected. The “@” prefix denotes a field that is reserved for a particular use. By default, most tools look for “@timestamp” with ElasticSearch.

The name of the host where this log message originated. In a Kubernetes cluster, this is the same as kubernetes.host.

The IPv4 address of the source server. Can be an array.

The IPv6 address of the source server, if available. Can be an array.

The logging level from various sources, including rsyslog(severitytext property), a Python logging module, and others.

The following values come from syslog.h, and are preceded by their numeric equivalents:

The two following values are not part of syslog.h but are widely used:

Map the log levels or priorities of other logging systems to their nearest match in the preceding list. For example, from python logging, you can match CRITICAL with crit, ERROR with err, and so on.

The process ID of the logging entity, if available.

The name of the service associated with the logging entity, if available. For example, syslog’s APP-NAME and rsyslog’s programname properties are mapped to the service field.