Configuring your logging deployment

Enabling multi-line exception detection

Enables multi-line error detection of container logs.

Log parsers often incorrectly identify separate lines of the same exception as separate exceptions. This leads to extra log entries and an incomplete or inaccurate view of the traced information.

Example java exception

java.lang.NullPointerException: Cannot invoke "String.toString()" because "<param1>" is null
    at testjava.Main.handle(Main.java:47)
    at testjava.Main.printMe(Main.java:19)
    at testjava.Main.main(Main.java:10)

• To enable logging to detect multi-line exceptions and reassemble them into a single log entry, ensure that the ClusterLogForwarder Custom Resource (CR) contains a detectMultilineErrors field, with a value of true.

Example ClusterLogForwarder CR

apiVersion: logging.openshift.io/v1
kind: ClusterLogForwarder
metadata:
  name: instance
  namespace: openshift-logging
spec:
  pipelines:
    - name: my-app-logs
      inputRefs:
        - application
      outputRefs:
        - default
      detectMultilineErrors: true

Details

When log messages appear as a consecutive sequence forming an exception stack trace, they are combined into a single, unified log record. The first log message’s content is replaced with the concatenated content of all the message fields in the sequence.

Troubleshooting

When enabled, the collector configuration will include a new section with type: detect_exceptions

Example vector configuration section

[transforms.detect_exceptions_app-logs]
 type = "detect_exceptions"
 inputs = ["application"]
 languages = ["All"]
 group_by = ["kubernetes.namespace_name","kubernetes.pod_name","kubernetes.container_name"]
 expire_after_ms = 2000
 multiline_flush_interval_ms = 1000

Example fluentd config section

<label @MULTILINE_APP_LOGS>
  <match kubernetes.**>
    @type detect_exceptions
    remove_tag_prefix 'kubernetes'
    message message
    force_line_breaks true
    multiline_flush_interval .2
  </match>
</label>