Log and trace configuration
Open Liberty has a unified logging component that handles messages that are written by applications and the runtime, and provides First Failure Data Capture (FFDC) capability. Logging data that is written by applications by using the System.out
, System.err
, or java.util.logging.Logger
streams is combined into the server logs.
A server has the following four primary log files:
console.log
- This file is created by theserver start
command. It contains the redirected standard output and standard error streams from the underlying JVM process. This console output is formatted for human readability and it lacks some information that is useful for automated log analysis. Theconsole.log
file is reset when the server is restarted.messages.log
- This file contains all messages that are written or captured by the logging component. All messages that are written to this file contain additional information such as the message timestamp and the ID of the thread that wrote the message. This file is suitable for automated log analysis. This file does not contain messages that are written directly by the JVM process.trace.log
- This file is created only if you enable trace. It contains all the content of themessages.log
file plus any enabled trace. This file does not contain messages that are written directly by the JVM process.verbosegc.XXX.log
- These files contain verbose garbage collection output from the JVM and are created by default when the Java implementation is either IBM Java or IBM Semeru Runtimes. Up to 10 rolling log files are created in your log directory, with 1024 GC cycles per file, whereXXX
represents the sequence number of the log file. Custom configurations take precedence over the default behavior. To disable the verbose garbage collection logs, addVERBOSEGC=false
to theserver.env
file. For more information, see the Verbose garbage collection logs.
You can use the user.timezone
JVM property to set the time zone for your application environments. Set the user.timezone
property in the jvm.properties
file. The updated time zone information is displayed in the messages.log
and trace.log
files. The following example shows how to set the time zone.
-Duser.timezone=time_zone_code
The following sections provide more information about configuring your Open Liberty logs:
Logging configuration
The logging component can be controlled through the server configuration. The logging component can be fully configured in your server.xml
file by the logging
element. However, logging is initialized before the server.xml
file is processed so configuring logging through the server.xml
file can result in early log entries that use a different log configuration from later ones. To avoid this problem, you can provide much of the logging configuration in the boostrap.properties
file and in some cases by using environment variables. Avoid specifying different values for the same configuration property in both the bootstrap.properties and the server.xml file. For more information about logging configuration settings, see the Configuration settings by source table.
Log file storage management
The console.log
file is created by redirecting the process stdout
and stderr
streams to a file. As a result, Open Liberty is unable to offer the same level of management, like log rollover, as it offers for the messages.log
file. If you are concerned about the size of the console.log
file, you can set the console log to OFF
and use the messages.log
file instead. All messages that are written or captured by the logging component are written to the messages.log
file. You can configure file rollover in the message log.
However, when you set the console log level to OFF
, messages from the underlying JVM are still written to the console.log
file. The same principle applies to any messages that are written to the native standard out and standard error streams. The Open Liberty logging component does not capture or control the destination of messages that are written directly by the JVM process, such as -verbose:gc
output. These messages are routed to the file system according to the active JVM settings and they do not appear in the messages.log
file.
To prevent the logging component from writing to the console log and configure the messages.log
file to roll over three times at 100 Mb, use the following configuration:
com.ibm.ws.logging.max.file.size=100
com.ibm.ws.logging.max.files=3
com.ibm.ws.logging.console.log.level=OFF
com.ibm.ws.logging.copy.system.streams=false
Log formats
Each of the three Open Liberty primary log files has a default log format and several alternative formats that you can specify in your logging configuration. The format that you choose for each log depends on the needs of your application and your infrastructure. For example, you might choose to output your message log in JSON format if it is consumed by a log analysis platform. Alternatively, you might choose the TBASIC
format for all three logs if you need a consistent log format across all of your log files.
The default format for the
console.log
file isDEV
.The default format for the
messages.log
file isSIMPLE
.The default format for the
trace.log
file isENHANCED
.
The following examples show sample output in each of the available log formats and list the logs for which each format is available: DEV
, SIMPLE
, JSON
, TBASIC
, ENHANCED
, and ADVANCED
.
- DEV (console log)
Messages that are in
DEV
format use the following pattern:[log level] <message_ID>: <message>
The following example shows a message in
DEV
format:[AUDIT] CWWKF0011I: The defaultServer server is ready to run a smarter planet. The defaultServer server started in 7.967 seconds.
- SIMPLE (console and message logs)
Messages that are in
SIMPLE
format use the following pattern:[time stamp] <thread ID> <fully qualified className> <log level attribute> <message ID>: <message>
The following example shows a message in
SIMPLE
format:[18/10/21 14:49:26:246 EDT] 0000003e com.ibm.ws.kernel.feature.internal.FeatureManager A CWWKF0011I: The defaultServer server is ready to run a smarter planet. The defaultServer server started in 7.844 seconds
In this example, the
A
attribute that precedes theCWWKF0011I
message ID code indicates that the message log level isAUDIT
. For more information, see Message log-level attributes.- JSON (console and message logs)
Messages in JSON format include field names that identify the different components of the message. The following example shows a message in
JSON
format:{"type":"liberty_message","host":"HOST","ibm_userDir":"/wlp\/usr\/","ibm_serverName":"defaultServer","message":"CWWKF0011I: The defaultServer server is ready to run a smarter planet. The defaultServer server started in 7.967 seconds.","ibm_threadId":"0000003e","ibm_datetime":"2021-10-18T14:50:58.159-0400","ibm_messageId":"CWWKF0011I","module":"com.ibm.ws.kernel.feature.internal.FeatureManager","loglevel":"AUDIT","ibm_sequence":"1634583058159_0000000000009"}
For more information, see JSON logging.
- TBASIC (console, messages, and trace logs)
Messages that are in
TBASIC
format use the following pattern:[time stamp] <thread ID> <fully qualified className> <log level attribute> <message ID>: <message>
The following example shows a message in
TBASIC
format:[18/10/21 14:47:19:718 EDT] 0000003e FeatureManage A CWWKF0011I: The defaultServer server is ready to run a smarter planet. The defaultServer server started in 7.718 seconds.
In this example, the
A
attribute that precedes theCWWKF0011I
message ID code indicates that the message log level isAUDIT
. For more information about log-level attributes, see Message log-level attributes.For more information about
TBASIC
, see TBASIC log format.- ENHANCED (trace log)
Messages in
ENHANCED
format include field names that are components from the log record for the particular trace message, which provide more details about the trace message. These following details are included:message type
date and time
messageID
fully qualified class name
severity level
log level
method name
class name
level value
thread name
correlationID (or product, component, message, and objectID)
The following example shows a message in
ENHANCED
format:[26/10/21 10:42:42:300 EDT] 0000006d id=00000000 x.com.ibm.ws.collector.manager.buffer.BufferManagerImpl 3 Adding event to buffer GenericData [type=com.ibm.ws.logging.source.trace,ibm_datetime=1635259362300,ibm_messageId=null,ibm_threadId=109,module=com.ibm.ws.event.internal.EventImpl,severity= <,loglevel=EXIT,ibm_methodName=setProperty,ibm_className=null,levelValue=400,threadName=Framework stop - Equinox Container: 2005505a-a1d7-46dc-abdd- 7230d055f81e,correlationId=,org=,product=,component=,message=Exit ,objectId=-964551509]
- ADVANCED (trace log)
Messages in
ADVANCED
format include field names and details that are similar toENHANCED
format, but that preserve more specific trace information.The following example shows a message in
ADVANCED
format:[26/10/21 10:29:25:011 EDT] 00000073 > UOW= source=com.ibm.ws.event.internal.EventImpl method=setProperty id=5f05cd40 org= prod= component= Entry service.objectClass java.util.concurrent.ExecutorService com.ibm.wsspi.threading.WSExecutorService
Message log-level attributes
Messages that are in DEV
format print the full log-level name before the message ID code. Messages that are in JSON
or ENHANCED
format include a loglevel
property that identifies the message log level.
Messages that are in SIMPLE
or TBASIC
format include one of the following message log-level attributes directly before the message ID code:
Attribute | Log level | Description |
---|---|---|
| AUDIT | Audit messages are written to the system output stream. |
| ERROR | Error messages are written to the system error stream. |
| INFO | Info messages are written to the system output stream. |
| WARNING | Warning messages are written to the system output stream. |
| N/A | This attribute does not specify a log level but indicates that the message is written to the system output stream. |
| N/A | This attribute does not specify a log level but indicates that the message is written to the system error stream. |
For more information about log levels and output streams, see the descriptions for the consoleLogLevel
and copySystemStreams
attributes for the logging element.
TBASIC log format
The TBASIC
logging format provides a consistent log format across your Open Liberty console, message, and trace log files. The BASIC
logging format exists for the trace log. However, the output differs between the trace log and that of the console and message logs. The TBASIC
format provides a consistent format option for the trace, console, and message logs that matches the BASIC
option for the trace log. The TBASIC
format acts as an alias for the BASIC
option.
You can specify the TBASIC
format in the bootstrap.properties
file, as shown in the following example:
com.ibm.ws.logging.message.format=tbasic com.ibm.ws.logging.console.format=tbasic com.ibm.ws.logging.trace.format=tbasic
You can also specify the TBASIC
log format for the messages and console logs as an environment variable in your server.env
file, as shown in the following example:
WLP_LOGGING_MESSAGE_FORMAT=TBASIC WLP_LOGGING_CONSOLE_FORMAT=TBASIC
JSON logging
You can simplify log parsing by producing your logs in JSON format. JSON is a self-describing format that many log analysis tools can consume without requiring format-specific parsing instructions. You can configure Open Liberty logs to produce logs in JSON format either by editing the bootstrap.properties
file or by specifying an environment variable. The following two examples show the configuration for each of these options:
Configure JSON logging in the
bootstrap.properties
file:com.ibm.ws.logging.message.format=json com.ibm.ws.logging.message.source=message,trace,accessLog,ffdc,audit
Configure JSON logging with environment variables in the
server.env
file:WLP_LOGGING_MESSAGE_FORMAT=json WLP_LOGGING_MESSAGE_SOURCE=message,trace,accessLog,ffdc,audit
Configurable JSON field names
When logs are in JSON format, you can use the jsonFieldMappings
attribute to replace default field names with new field names. Replacing the default field names might be necessary if other servers in the same logging configuration use different field names than the Open Liberty default names. For example, an Open Liberty message is referred to by the message
field name, but the message in another container might be in a field called log
. In this case, two different visualizations of the messages show in the logs on a dashboard. If you modify the Open Liberty output field name so that it matches the other log, you can view them in the same visualization. The following examples show sample configurations for renaming a JSON field.
To configure a new field name, you can include the following environmental variable in the
server.env
file:WLP_LOGGING_JSON_FIELD_MAPPINGS=loglevel:level
In this example, the
loglevel
field name is replaced by thelevel
field name.To configure a field name for a specific source, you can include the following environmental variable in the
server.env
file:WLP_LOGGING_JSON_FIELD_MAPPINGS=message:message:log
In this example, the
message
field name is replaced by thelog
field name in the message log.
To omit a field from the logs, specify the field name without a replacement, as shown in the following example:
WLP_LOGGING_JSON_FIELD_MAPPINGS=defaultFieldName:
To rename or omit multiple fields, specify a comma-separated list of field name mappings.
For a full list of the default JSON field names, see the JSON log events reference list.
For more information, see Configurable JSON log field names.
Custom message and trace JSON fields
You can add custom fields to your JSON-formatted message and trace output to gather information about a particular issue or incident. For example, if you want to check the requests from a specific user, you can add a custom field to filter application logs by that user’s ID. You can add another field for the session ID so that you can analyze and filter application logs for a specific session.
The Open Liberty LogRecordContext
API can add custom fields to your log and trace records. This API adds a field by specifying a value for an extension. To use the LogRecordContext
API, first import the com.ibm.websphere.logging.hpel.LogRecordContext
class. The following examples show how to add different kinds of custom fields to your JSON logs.
To add a string-valued field to your application logs, you can include the following line in your application:
LogRecordContext.addExtension("userName","bob");
The newly specified field is added to log and trace entries that are created on the same thread that executes the addExtension
method. In this example, a custom field that is called userName
is added for the bob
user ID.
To add fields with boolean, float, int, or long values, the extension name must include the suffixes _bool
, _float
, _int
, or _long
, as shown in the following examples:
LogRecordContext.addExtension("extensionName_bool","true");
LogRecordContext.addExtension("extensionName_int","112233");
LogRecordContext.addExtension("extensionName_float","1.2");
LogRecordContext.addExtension("extensionName_long","132");
When you specify these suffixes to add non-string values, the resulting JSON field values are not enclosed in quotes in the logs. Only string-valued JSON field values are enclosed in quotes in the logs.
To remove custom fields from the logs, use the following method:
LogRecordContext.removeExtension(extensionName);
After you remove an extension, JSON output for subsequent logs and trace that are made on the same thread do not include that field.
Selectable HTTP access log JSON fields
When logs are in JSON format and the accessLog
source is specified, you can replace the default HTTP access log JSON fields with a different set of fields. You can use the jsonAccessLogFields
attribute to specify whether your access logs use the default fields or a set of fields that is specified by the logFormat
attribute. You specify the replacement fields in the logFormat
attribute of the accessLogging
element within the httpEndpoint
element.
With this configuration, you can receive information that is otherwise not available in JSON logs, such as the remote user ID, request headers, and more. These logs can be used by log analysis tools, such as the Elastic stack, to monitor your server. For more information about HTTP access log format options, see HTTP access logging.
The following example shows a configuration in the server.xml
file to replace the default HTTP access log fields with fields that are specified by the logFormat
attribute. HTTP access logging must be enabled to receive JSON access logs.
<httpEndpoint httpPort="9080" httpsPort="9443" id="defaultHttpEndpoint">
<accessLogging logFormat='%R{W} %u %{my_cookie}C %s'/>
</httpEndpoint>
<logging jsonAccessLogFields="logFormat"/>
Container image logging configuration
In containerized environments, you can disable the message log and format the console output as JSON by using environment variables, as shown in the following example:
WLP_LOGGING_MESSAGE_FORMAT=json
WLP_LOGGING_MESSAGE_SOURCE=
WLP_LOGGING_CONSOLE_FORMAT=json
WLP_LOGGING_CONSOLE_LOGLEVEL=info
WLP_LOGGING_CONSOLE_SOURCE=message,trace,accessLog,ffdc,audit
You can specify this configuration when you run the docker run
command by using the -e
flag to set the environment variables:
docker run -e "WLP_LOGGING_CONSOLE_SOURCE=message,trace,accessLog,ffdc"
-e "WLP_LOGGING_CONSOLE_FORMAT=json"
-e "WLP_LOGGING_CONSOLE_LOGLEVEL=info"
-e "WLP_LOGGING_MESSAGE_FORMAT=json"
-e "WLP_LOGGING_MESSAGE_SOURCE=" open-liberty
If you use Podman to manage your containers, run the following command:
podman run -e "WLP_LOGGING_CONSOLE_SOURCE=message,trace,accessLog,ffdc"
-e "WLP_LOGGING_CONSOLE_FORMAT=json"
-e "WLP_LOGGING_CONSOLE_LOGLEVEL=info"
-e "WLP_LOGGING_MESSAGE_FORMAT=json"
-e "WLP_LOGGING_MESSAGE_SOURCE=" open-liberty
Binary logging
Liberty has a high-performance binary log format option that reduces the resources that are needed to write trace files. For best performance, when you configure binary logging, set the console log level to OFF
and use the message log instead. You can configure binary logging in the bootstrap.properties
file, as shown in the following example:
websphere.log.provider=binaryLogging-1.0
com.ibm.ws.logging.console.log.level=OFF
com.ibm.ws.logging.copy.system.streams=false
However, when you set the console log level to OFF
, messages from the underlying JVM other other messages that are written to the native standard out and standard error streams are still written to the console.log
file.
You can use the binaryLog
command to convert the binary log to a text file, as shown in the following example for a server that is called defaultServer
:
binaryLog view defaultServer
Time-based log rollover
You can enable time-based periodic rollover of Liberty messages and trace logs by specifying a log rollover start time and a log rollover interval duration. The specified rollover start time is the time of day when logs first roll over. The rollover interval duration is the time interval in between consecutive log rollovers.
For example, a server with a rollover start time of midnight and a rollover interval of 1 day rolls over the messages and trace logs once every day, at midnight.
You can configure rollover for Open Liberty messages and trace logs by either editing the bootstrap.properties
file, specifying environment variables, or by using server configuration attributes. The following examples show the configuration for each of these options, with a log rollover start time of midnight and a log rollover interval of 1 day.
The following example configures time-based log rollover in the
bootstrap.properties
file.com.ibm.ws.logging.rollover.start.time=00:00 com.ibm.ws.logging.rollover.interval=1m
The following example configures time-based log rollover with environment variables in the
server.env
file.WLP_LOGGING_ROLLOVER_START_TIME=00:00 WLP_LOGGING_ROLLOVER_INTERVAL=1m
The following example configures time-based log rollover with server configuration attributes in the
server.xml
file.<logging rolloverStartTime="00:00" rolloverInterval="1d"/>
Configuration settings by source
The following table shows the equivalent server.xml
file, bootstrap.properties
file, and environment variable configurations along with brief descriptions. Settings that apply to the console log appear first, followed by settings that apply to the message log, and then the trace log. Settings that apply either globally or to more than one log type appear last in the table. For more information, see the logging element.
Server XML Attribute | bootstrap property | Env var | Description |
---|---|---|---|
consoleFormat | com.ibm.ws.logging.console.format | WLP_LOGGING_CONSOLE_FORMAT | This setting specifies the required format for the console. Valid values are |
consoleLogLevel | com.ibm.ws.logging.console.log.level | WLP_LOGGING_CONSOLE_LOGLEVEL | This setting controls the granularity of messages that go to the console. The valid values are INFO, AUDIT, WARNING, ERROR, and OFF. The default is AUDIT. If using with the Eclipse developer tools this must be set to the default. |
consoleSource | com.ibm.ws.logging.console.source | WLP_LOGGING_CONSOLE_SOURCE | This setting specifies a comma-separated list of sources that route to the console. It applies only when the console format is set to |
copySystemStreams | com.ibm.ws.logging.copy.system.streams | If this setting is set to | |
messageFileName | com.ibm.ws.logging.message.file.name | This setting specifies the name of the message log file. The message log file has a default name of | |
messageFormat | com.ibm.ws.logging.message.format | WLP_LOGGING_MESSAGE_FORMAT | This setting specifies the required format for the |
messageSource | com.ibm.ws.logging.message.source | WLP_LOGGING_MESSAGE_SOURCE | This setting specifies a list of comma-separated sources that route to the |
suppressSensitiveTrace | This attribute, when set to | ||
traceFileName | com.ibm.ws.logging.trace.file.name | This setting specifies the name of the trace log file. The | |
traceFormat | com.ibm.ws.logging.trace.format | This setting controls the format of the trace log. The default format for Liberty is | |
traceSpecification | com.ibm.ws.logging.trace.specification | This setting is used to selectively enable trace. The log detail level specification is in the following format:
The component specifies what log sources the level is set for. A component can be a logger name, trace group, or class name. The level specifies what level of trace to output for that component by using one of the following levels:
You can provide multiple log detail level specifications that are separated by colons. An asterisk * acts as a wildcard to match multiple components based on a prefix. For example: - - - The | |
com.ibm.ws.logging.newLogsOnStart | If this setting is set to true when Open Liberty starts, any existing | ||
hideMessage | com.ibm.ws.logging.hideMessage | You can use this setting to configure the messages keys that you want to hide from the | |
isoDateFormat | com.ibm.ws.logging.isoDateFormat | This setting specifies whether to use ISO-8601 formatted dates in log files. The default value is false. If this setting is set to true, the ISO-8601 format is used in the If you specify a value of | |
maxFfdcAge | com.ibm.ws.logging.max.ffdc.age | WLP_LOGGING_MAX_FFDC_AGE | Use this setting to enable FFDC file purging based on age. This setting specifies the maximum file age before an FFDC file is deleted. Purges occur everyday at midnight. Specify a positive integer followed by a unit of time, which can be days ( |
appsWriteJson | com.ibm.ws.logging.apps.write.json | WLP_LOGGING_APPS_WRITE_JSON | When the message log or console is in JSON format, this setting allows applications to write JSON-formatted messages to those destinations, without modification. |
jsonAccessLogFields | com.ibm.ws.json.access.log.fields | WLP_LOGGING_JSON_ACCESS_LOG_FIELDS | When logs are in JSON format, you can use this setting to replace the default HTTP access log JSON fields with fields that are specified by the |
jsonFieldMappings | com.ibm.ws.logging.json.field.mappings | WLP_LOGGING_JSON_FIELD_MAPPINGS | When logs are in JSON format, use this setting to replace default field names with new field names or to omit fields from the logs. For more information, see Configurable JSON field names |
logDirectory | com.ibm.ws.logging.log.directory | LOG_DIR | You can use this setting to set a directory for all log files, excluding the |
maxFiles | com.ibm.ws.logging.max.files | This setting specifies how many of each of the logs files are kept. This setting also applies to the number of exception summary logs for FFDC. So if this number is | |
maxFileSize | com.ibm.ws.logging.max.file.size | This setting specifies the maximum size (in MB) that a log file can reach before it is rolled. Setting the value to | |
stackTraceSingleEntry | com.ibm.ws.logging.stackTraceSingleEntry | WLP_LOGGING_STACK_TRACE_SINGLE_ENTRY | When this attribute is set to |
rolloverStartTime | com.ibm.ws.logging.rollover.start.time | WLP_LOGGING_ROLLOVER_START_TIME | Use this setting alone or with the |
rolloverInterval | com.ibm.ws.logging.rollover.interval | WLP_LOGGING_ROLLOVER_INTERVAL | Use this setting alone or with the |
Trace log detail levels
You can specify the traceSpecification
attribute for the logging element in your server.xml
file to selectively enable trace and set the trace log detail level for specific components in your server configuration. The following example specifies a trace specification for two groups of components.
<logging traceSpecification="com.myco.mypackage1.*=fine:com.myco.mypackage2.*=finest"/>
In this example, the two detail level specifications are separated by a colon (:
).
The com.myco.mypackage1.*=fine
value applies to all classes with the package name that begins with com.myco.mypackage1
.
The com.myco.mypackage2.*=finest
value applies to all classes with the package name that begins with com.myco.mypackage2
.
The following table lists the valid log detail levels that you can specify in the traceSpecification
attribute. You can specify an empty value for any component to disable all trace for that component. Any component that is not specified is initialized to a default state of *=info
.
Setting a trace component to off
, fatal
, severe
, warning
, audit
, or info
level has the same effect as setting it to the info
level as info
and higher levels cannot be disabled. The trace.log
file is created and logs are written to it only if one or more trace components are set to config
level or lower.
Logging level | Content or Significance |
---|---|
off | Logging is turned off. |
fatal | Task cannot continue and component, application, and server cannot function. |
severe | Task cannot continue but component, application, and server can still function. This level can also indicate an impending unrecoverable error. |
warning | Potential error or impending error. This level can also indicate a progressive failure. For example, the potential leaking of resources. |
audit | Significant event that affects server state or resources |
info | General information that outlines overall task progress |
config | Configuration change or status |
detail | General information that details subtask progress |
fine | General trace |
finer | Detailed trace that includes method entry, exit, and return values |
finest | A more detailed trace that includes all the detail that is needed to debug problems |
all | All events are logged. If you create custom levels, |