Structured logging with the legacy Logging agent

This document discusses the concept of structured logging and the how to add structure to log entry payload fields using the legacy Logging agent. This information applies to Compute Engine VMs; it doesn't apply to other services such as Google Kubernetes Engine, App Engine flexible environment, and Cloud Run functions.

When the log payload is formatted as a JSON object and that object is stored in the jsonPayload field, the log entry is called a structured log. For these logs, you can construct queries that search specific JSON paths and you can index specific fields in the log payload. In contrast, when the log payload is formatted as a string and stored in the textPayload field, the log entry is unstructured. You can search the text field, but you can't index its content.

Write logs by using the legacy Logging agent

To create structured log entries, you can install and configure the legacy Logging agent to collect logs from your Compute Engine instances and send them to Cloud Logging.

Special JSON fields

Some fields in the JSON object are recognized as special by the legacy Logging agent and extracted into the LogEntry structure. These special JSON fields can be used to set the following fields in the LogEntry:

• severity
• spanId
• labels defined by the user
• httpRequest

Because JSON is more precise and versatile than text lines, you can use JSON objects to write multiline messages and add metadata.

To create structured log entries for your applications using the simplified format, see the following table, which lists the fields and their values in JSON:

To create log entries in the simplified format, create a JSON representation of the entry using the fields. All of the fields are optional.

The following is an example of a simplified JSON log entry:

{
  "severity":"ERROR",
  "message":"There was an error in the application.",
  "httpRequest":{
    "requestMethod":"GET"
  },
  "time":"2020-10-12T07:20:50.52Z",
  "logging.googleapis.com/insertId":"42",
  "logging.googleapis.com/labels":{
    "user_label_1":"value_1",
    "user_label_2":"value_2"
  },
  "logging.googleapis.com/operation":{
    "id":"get_data",
    "producer":"github.com/MyProject/MyApplication",
    "first":"true"
  },
  "logging.googleapis.com/sourceLocation":{
    "file":"get_data.py",
    "line":"142",
    "function":"getData"
  },
  "logging.googleapis.com/spanId":"000000000000004a",
  "logging.googleapis.com/trace":"projects/my-projectid/traces/06796866738c859f2f19b7cfb3214824",
  "logging.googleapis.com/trace_sampled":false
}

The following is an example of the resulting log entry:

{
  "insertId": "42",
  "jsonPayload": {
    "message": "There was an error in the application",
    "time": "2020-10-12T07:20:50.52Z"
  },
  "httpRequest": {
    "requestMethod": "GET"
  },
  "resource": {
    "type": "k8s_container",
    "labels": {
      "container_name": "hello-app",
      "pod_name": "helloworld-gke-6cfd6f4599-9wff8",
      "project_id": "stackdriver-sandbox-92334288",
      "namespace_name": "default",
      "location": "us-west4",
      "cluster_name": "helloworld-gke"
    }
  },
  "timestamp": "2020-10-12T07:20:50.52Z",
  "severity": "ERROR",
  "labels": {
    "user_label_2": "value_2",
    "user_label_1": "value_1"
  },
  "logName": "projects/stackdriver-sandbox-92334288/logs/stdout",
  "operation": {
    "id": "get_data",
    "producer": "github.com/MyProject/MyApplication",
    "first": true
  },
  "trace": "projects/my-projectid/traces/06796866738c859f2f19b7cfb3214824",
  "sourceLocation": {
    "file": "get_data.py",
    "line": "142",
    "function": "getData"
  },
  "receiveTimestamp": "2020-10-12T07:20:57.52Z",
  "spanId": "000000000000004a"
}

Configuration

The legacy Logging agent, google-fluentd, is a Cloud Logging-specific packaging of the Fluentd log data collector. The Logging agent comes with the default Fluentd configuration and uses Fluentd input plugins to pull event logs from external sources such as files on disk, or to parse incoming log records.

Fluentd has a list of supported parsers that extract logs and convert them into structured (JSON) payloads.

By configuring a log source with format [PARSER_NAME], you can build on the built-in parsers provided by Fluentd. For information about configuring the legacy Logging agent, see Configure the Logging agent.

The following code samples show the Fluentd configuration, the input log record, and the output structured payload, which is part of a Cloud Logging log entry:

• Fluentd configuration:

    <source>
      @type tail
  
      format syslog # This uses a predefined log format regex named
                    # `syslog`. See details at https://docs.fluentd.org/parser/syslog.
  
      path /var/log/syslog
      pos_file /var/lib/google-fluentd/pos/syslog.pos
      read_from_head true
      tag syslog
    </source>
  

• Log record (input):

    <6>Feb 28 12:00:00 192.168.0.1 fluentd[11111]: [error] Syslog test
  

• Structured payload (output):

    jsonPayload: {
        "pri": "6",
        "host": "192.168.0.1",
        "ident": "fluentd",
        "pid": "11111",
        "message": "[error] Syslog test"
    }
  

For more information about how the syslog parser works, see the detailed Fluentd documentation.

Standard parsers enabled by default

The following table includes the standard parsers that are included in the agent if you enable structured logging:

For instructions on enabling structured logging when installing the legacy Logging agent, see the Installation section.

Installation

To enable structured logging, you must change the default configuration of the legacy Logging agent when installing or reinstalling it. Enabling structured logging replaces the previously listed configuration files but doesn't change the operation of the agent itself.

When you enable structured logging, the listed logs are converted to log entries with different formats than they had before you enabled structured logs. If the logs are being routed to destinations outside of Logging, the change could affect any post-processing applications. For example, if routing logs to BigQuery, BigQuery rejects the new log entries for the remainder of the day as having an incorrect schema.

For instructions on installing the legacy Logging agent and enabling structured logging, refer to Installing the Logging agent.

You can find the legacy Logging agent configuration files at /etc/google-fluentd/config.d/, which should now include the Standard parsers enabled by default.

Configure Apache access log format

By default, the legacy Logging agent stores Apache access log data in the jsonPayload field. For example:

{
  "logName": ...,
  "resource": ...,
  "httpRequest": ...,
  "jsonPayload": {
    "user"   : "some-user",
    "method" : "GET",
    "code"   : 200,
    "size"   : 777,
    "host"   : "192.168.0.1",
    "path"   : "/some-path",
    "referer": "some-referer",
    "agent"  : "Opera/12.0"
  },
  ...
}

Alternatively, you can configure the legacy Logging agent to extract certain fields to the httpRequest field. For example:

{
  "logName": ...,
  "resource": ...,
  "httpRequest": {
    "requestMethod": "GET",
    "requestUrl": "/some-path",
    "requestSize": "777",
    "status": "200",
    "userAgent": "Opera/12.0",
    "serverIp": "192.168.0.1",
    "referrer":"some-referrer",
  },
  "jsonPayload": {
    "user":"some-user"
  },
  ...
}

Configuring the httpRequest field, as shown in the prior sample, assists tracing: the Google Cloud console presents all logs for a given HTTP request in a parent-child hierarchy.

To configure this extraction, add the following to the end of your /etc/google-fluentd/config.d/apache.conf:

<filter apache-access>
  @type record_transformer
  enable_ruby true
  <record>
    httpRequest ${ {"requestMethod" => record['method'], "requestUrl" => record['path'], "requestSize" => record['size'], "status" => record['code'], "userAgent" => record['agent'], "serverIp" => record['host'],
    "referer" => record['referer']} }
  </record>
  remove_keys method, path, size, code, agent, host, referer
</filter>

For more details on how to configure your log entries, see Modifying log records.

Configure nginx access log format

By default, the legacy Logging agent stores nginx access log data in the jsonPayload field. For example:

{
  "logName": ...,
  "resource": ...,
  "httpRequest": ...,
  "jsonPayload": {
    "remote":"127.0.0.1",
    "host":"192.168.0.1",
    "user":"some-user",
    "method":"GET",
    "path":"/some-path",
    "code":"200",
    "size":"777",
    "referrer":"some-referrer",
    "agent":"Opera/12.0",
    "http_x_forwarded_for":"192.168.3.3"
  },
  ...
}

Alternatively, you can configure the legacy Logging agent to extract certain fields to the httpRequest field. For example:

{
  "logName": ...,
  "resource": ...,
  "httpRequest": {
    "requestMethod": "GET",
    "requestUrl": "/some-path",
    "requestSize": "777",
    "status": "200",
    "userAgent": "Opera/12.0",
    "remoteIp": "127.0.0.1",
    "serverIp": "192.168.0.1",
    "referrer":"some-referrer",
  },
  "jsonPayload": {
    "user":"some-user",
    "http_x_forwarded_for":"192.168.3.3"
  },
  ...
}

Configuring the httpRequest field, as shown in the prior sample, assists tracing: the Google Cloud console presents all logs for a given HTTP request in a parent-child hierarchy.

To configure this extraction, add the following to the end of your /etc/google-fluentd/config.d/nginx.conf:

<filter nginx-access>
  @type record_transformer
  enable_ruby true
  <record>
    httpRequest ${ {"requestMethod" => record['method'], "requestUrl" => record['path'], "requestSize" => record['size'], "status" => record['code'], "userAgent" => record['agent'], "remoteIp" => record['remote'], "serverIp" => record['host'], "referer" => record['referer']} }
  </record>
  remove_keys method, path, size, code, agent, remote, host, referer
</filter>

For more details on how to configure your log entries, see Modifying log records.

Write your own parser

If your logs aren't supported by the standard parsers, you can write your own parser. Parsers consist of a regular expression that is used to match log records and apply labels to the pieces.

The following code examples show a log line in the log record, a configuration with a regular expression that indicates the format of the log line, and the stored log entry:

• A log line in the log record:

  REPAIR CAR $500
  

• A configuration with a regular expression that indicates the format of the log line:

  $ sudo vim /etc/google-fluentd/config.d/test-structured-log.conf
  $ cat /etc/google-fluentd/config.d/test-structured-log.conf
  <source>
    @type tail
  
    # Format indicates the log should be translated from text to
    # structured (JSON) with three fields, "action", "thing" and "cost",
    # using the following regex:
    format /(?<action>\w+) (?<thing>\w+) \$(?<cost>\d+)/
    # The path of the log file.
    path /tmp/test-structured-log.log
    # The path of the position file that records where in the log file
    # we have processed already. This is useful when the agent
    # restarts.
    pos_file /var/lib/google-fluentd/pos/test-structured-log.pos
    read_from_head true
    # The log tag for this log input.
    tag structured-log
  </source>
  

• The resulting log entry:

  {
  insertId:  "eps2n7g1hq99qp"
  jsonPayload: {
    "action": "REPAIR"
    "thing": "CAR"
    "cost": "500"
  }
  labels: {
    compute.googleapis.com/resource_name:  "add-structured-log-resource"
  }
  logName:  "projects/my-sample-project-12345/logs/structured-log"
  receiveTimestamp:  "2023-03-21T01:47:11.475065313Z"
  resource: {
    labels: {
      instance_id:  "3914079432219560274"
      project_id:  "my-sample-project-12345"
      zone:  "us-central1-c"
    }
    type:  "gce_instance"
  }
  timestamp:  "2023-03-21T01:47:05.051902169Z"
  }
  

Troubleshoot issues

To troubleshoot common issues found with installing or interacting with the legacy Logging agent, see Troubleshooting the agent.