Logbook: HTTP request and response logging

Logbook noun, /lɑɡ bʊk/: A book in which measurements from the ship's log are recorded, along with other salient details of the voyage.

Logbook is an extensible Java library to enable complete request and response logging for different client- and server-side technologies. It satisfies a special need by a) allowing web application developers to log any HTTP traffic that an application receives or sends b) in a way that makes it easy to persist and analyze it later. This can be useful for traditional log analysis, meeting audit requirements or investigating individual historic traffic issues.

Logbook is ready to use out of the box for most common setups. Even for uncommon applications and technologies, it should be simple to implement the necessary interfaces to connect a library/framework/etc. to it.

Features

• Logging: of HTTP requests and responses, including the body; partial logging (no body) for unauthorized requests
• Customization: of logging format, logging destination, and conditions that request to log
• Support: for Servlet containers, Apache’s HTTP client, Square's OkHttp, and (via its elegant API) other frameworks
• Optional obfuscation of sensitive data
• Spring Boot Auto Configuration
• Scalyr compatible
• Sensible defaults

Dependencies

• Java 8
• Any build tool using Maven Central, or direct download
• Servlet Container (optional)
• Apache HTTP Client (optional)
• JAX-RS 2.x Client and Server (optional)
• Netty 4.x (optional)
• OkHttp 2.x or 3.x (optional)
• Spring 4.x or 5.x (optional)
• Spring Boot 1.x or 2.x (optional)
• logstash-logback-encoder 5.x (optional)

Installation

Add the following dependency to your project:

<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-core</artifactId>
    <version>${logbook.version}</version>
</dependency>

Additional modules/artifacts of Logbook always share the same version number.

Alternatively, you can import our bill of materials...

<dependencyManagement>
  <dependencies>
    <dependency>
      <groupId>org.zalando</groupId>
      <artifactId>logbook-bom</artifactId>
      <version>${logbook.version}</version>
      <type>pom</type>
      <scope>import</scope>
    </dependency>
  </dependencies>
</dependencyManagement>

... which allows you to omit versions:

<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-core</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-httpclient</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-jaxrs</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-json</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-netty</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-okhttp</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-okhttp2</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-servlet</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-spring-boot-starter</artifactId>
</dependency>
<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-logstash</artifactId>
</dependency>

The logbook logger must be configured to trace level in order to log the requests and responses. With Spring Boot 2 (using Logback) this can be accomplised by adding the following line to your application.properties

logging.level.org.zalando.logbook: TRACE

Usage

All integrations require an instance of Logbook which holds all configuration and wires all necessary parts together. You can either create one using all the defaults:

Logbook logbook = Logbook.create();

or create a customized version using the LogbookBuilder:

Logbook logbook = Logbook.builder()
    .condition(new CustomCondition())
    .queryFilter(new CustomQueryFilter())
    .pathFilter(new CustomPathFilter())
    .headerFilter(new CustomHeaderFilter())
    .bodyFilter(new CustomBodyFilter())
    .requestFilter(new CustomRequestFilter())
    .responseFilter(new CustomResponseFilter())
    .sink(new DefaultSink(
            new CustomHttpLogFormatter(),
            new CustomHttpLogWriter()
    ))
    .build();

Strategy

Logbook used to have a very rigid strategy how to do request/response logging:

• Requests/responses are logged separately
• Requests/responses are logged soon as possible
• Requests/responses are logged as a pair or not logged at all
  (i.e. no partial logging of traffic)

Some of those restrictions could be mitigated with custom HttpLogWriter implementations, but they were never ideal.

Starting with version 2.0 Logbook now comes with a Strategy pattern at its core. Make sure you read the documentation of the Strategy interface to understand the implications.

Logbook comes with some built-in strategies:

• BodyOnlyIfStatusAtLeastStrategy
• StatusAtLeastStrategy
• WithoutBodyStrategy

Phases

Logbook works in several different phases:

1. Conditional,
2. Filtering,
3. Formatting and
4. Writing

Each phase is represented by one or more interfaces that can be used for customization. Every phase has a sensible default.

Conditional

Logging HTTP messages and including their bodies is a rather expensive task, so it makes a lot of sense to disable logging for certain requests. A common use case would be to ignore health check requests from a load balancer, or any request to management endpoints typically issued by developers.

Defining a condition is as easy as writing a special Predicate that decides whether a request (and its corresponding response) should be logged or not. Alternatively you can use and combine predefined predicates:

Logbook logbook = Logbook.builder()
    .condition(exclude(
        requestTo("/health"),
        requestTo("/admin/**"),
        contentType("application/octet-stream"),
        header("X-Secret", newHashSet("1", "true")::contains)))
    .build();

Exclusion patterns, e.g. /admin/**, are loosely following Ant's style of path patterns without taking the the query string of the URL into consideration.

Filtering

The goal of Filtering is to prevent the logging of certain sensitive parts of HTTP requests and responses. This usually includes the Authorization header, but could also apply to certain plaintext query or form parameters — e.g. password.

Logbook supports different types of filters:

QueryFilter, PathFilter, HeaderFilter and BodyFilter are relatively high-level and should cover all needs in ~90% of all cases. For more complicated setups one should fallback to the low-level variants, i.e. RequestFilter and ResponseFilter respectively (in conjunction with ForwardingHttpRequest/ForwardingHttpResponse).

You can configure filters like this:

Logbook logbook = Logbook.builder()
    .requestFilter(replaceBody(contentType("audio/*"), "mmh mmh mmh mmh"))
    .responseFilter(replaceBody(contentType("*/*-stream"), "It just keeps going and going..."))
    .queryFilter(accessToken())
    .queryFilter(replaceQuery("password", "<secret>"))
    .headerFilter(authorization()) 
    .headerFilter(eachHeader("X-Secret"::equalsIgnoreCase, "<secret>"))
    .build();

You can configure as many filters as you want - they will run consecutively.

Correlation

Logbook uses a correlation id to correlate requests and responses. This allows match-related requests and responses that would usually be located in different places in the log file.

If the default implementation of the correlation id is insufficient for your use case, you may provide a custom implementation:

Logbook logbook = Logbook.builder()
    .correlationId(new CustomCorrelationId())
    .build();

Formatting

Formatting defines how requests and responses will be transformed to strings basically. Formatters do not specify where requests and responses are logged to — writers do that work.

Logbook comes with two different default formatters: HTTP and JSON.

HTTP

HTTP is the default formatting style, provided by the DefaultHttpLogFormatter. It is primarily designed to be used for local development and debugging, not for production use. This is because it’s not as readily machine-readable as JSON.

Request

Incoming Request: 2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b
GET http://example.org/test HTTP/1.1
Accept: application/json
Host: localhost
Content-Type: text/plain

Hello world!

Response

Outgoing Response: 2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b
Duration: 25 ms
HTTP/1.1 200
Content-Type: application/json

{"value":"Hello world!"}

JSON

JSON is an alternative formatting style, provided by the JsonHttpLogFormatter. Unlike HTTP, it is primarily designed for production use — parsers and log consumers can easily consume it.

Requires the following dependency:

<dependency>
  <groupId>org.zalando</groupId>
  <artifactId>logbook-json</artifactId>
</dependency>

Request

{
  "origin": "remote",
  "type": "request",
  "correlation": "2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b",
  "protocol": "HTTP/1.1",
  "sender": "127.0.0.1",
  "method": "GET",
  "uri": "http://example.org/test",
  "host": "example.org",
  "path": "/test",
  "scheme": "http",
  "port": null,
  "headers": {
    "Accept": ["application/json"],
    "Content-Type": ["text/plain"]
  },
  "body": "Hello world!"
}

Response

{
  "origin": "local",
  "type": "response",
  "correlation": "2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b",
  "duration": 25,
  "protocol": "HTTP/1.1",
  "status": 200,
  "headers": {
    "Content-Type": ["text/plain"]
  },
  "body": "Hello world!"
}

Note: Bodies of type application/json (and application/*+json) will be inlined into the resulting JSON tree. I.e., a JSON response body will not be escaped and represented as a string:

{
  "origin": "local",
  "type": "response",
  "correlation": "2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b",
  "duration": 25,
  "protocol": "HTTP/1.1",
  "status": 200,
  "headers": {
    "Content-Type": ["application/json"]
  },
  "body": {
    "greeting": "Hello, world!"
  }
}

Common Log Format

The Common Log Format (CLF) is a standardized text file format used by web servers when generating server log files. The format is supported via the CommonLogFormatSink:

185.85.220.253 - - [02/Aug/2019:08:16:41 0000] "GET /search?q=zalando HTTP/1.1" 200 -

cURL

cURL is an alternative formatting style, provided by the CurlHttpLogFormatter which will render requests as executable cURL commands. Unlike JSON, it is primarily designed for humans.

Request

curl -v -X GET 'http://localhost/test' -H 'Accept: application/json'

Response

See HTTP or provide own fallback for responses:

new CurlHttpLogFormatter(new JsonHttpLogFormatter());

Splunk

Splunk is an alternative formatting style, provided by the SplunkHttpLogFormatter which will render requests and response as key-value pairs.

Request

origin=remote type=request correlation=2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b protocol=HTTP/1.1 sender=127.0.0.1 method=POST uri=http://example.org/test host=example.org scheme=http port=null path=/test headers={Accept=[application/json], Content-Type=[text/plain]} body=Hello world!

Response

origin=local type=response correlation=2d66e4bc-9a0d-11e5-a84c-1f39510f0d6b duration=25 protocol=HTTP/1.1 status=200 headers={Content-Type=[text/plain]} body=Hello world!

Writing

Writing defines where formatted requests and responses are written to. Logbook comes with three implementations: Logger, Stream and Chunking.

Logger

By default, requests and responses are logged with an slf4j logger that uses the org.zalando.logbook.Logbook category and the log level trace. This can be customized:

Logbook logbook = Logbook.builder()
    .sink(new DefaultSink(
            new DefaultHttpFormatter(),
            new DefaultHttpLogWriter())
    .build();

Stream

An alternative implementation is to log requests and responses to a PrintStream, e.g. System.out or System.err. This is usually a bad choice for running in production, but can sometimes be useful for short-term local development and/or investigation.

Logbook logbook = Logbook.builder()
    .sink(new DefaultSink(
            new DefaultHttpFormatter(),
            new StreamHttpLogWriter(System.err)
    ))
    .build();

Chunking

The ChunkingSink will split long messages into smaller chunks and will write them individually while delegating to another sink:

Logbook logbook = Logbook.builder()
    .sink(new ChunkingSink(sink, 1000))
    .build();

Sink

The combination of HttpLogFormatter and HttpLogWriter suits most use cases well, but it has limitations. Implementing the Sink interface directly allows for more sophisticated use cases, e.g. writing requests/responses to a structured persistent storage like a database.

Multiple sinks can be combined into one using the CompositeSink.

Servlet

You’ll have to register the LogbookFilter as a Filter in your filter chain — either in your web.xml file (please note that the xml approach will use all the defaults and is not configurable):

<filter>
    <filter-name>LogbookFilter</filter-name>
    <filter-class>org.zalando.logbook.servlet.LogbookFilter</filter-class>
</filter>
<filter-mapping>
    <filter-name>LogbookFilter</filter-name>
    <url-pattern>/*</url-pattern>
    <dispatcher>REQUEST</dispatcher>
    <dispatcher>ASYNC</dispatcher>
</filter-mapping>

or programmatically, via the ServletContext:

context.addFilter("LogbookFilter", new LogbookFilter(logbook))
    .addMappingForUrlPatterns(EnumSet.of(REQUEST, ASYNC), true, "/*"); 

Beware: The ERROR dispatch is not supported. You're strongly advised to produce error responses within the REQUEST or ASNYC dispatch.

The LogbookFilter will, by default, treat requests with a application/x-www-form-urlencoded body not different from any other request, i.e you will see the request body in the logs. The downside of this approach is that you won't be able to use any of the HttpServletRequest.getParameter*(..) methods. See issue #94 for some more details.

Form Requests

As of Logbook 1.5.0, you can now specify one of three strategies that define how Logbook deals with this situation by using the logbook.servlet.form-request system property:

Security

Secure applications usually need a slightly different setup. You should generally avoid logging unauthorized requests, especially the body, because it quickly allows attackers to flood your logfile — and, consequently, your precious disk space. Assuming that your application handles authorization inside another filter, you have two choices:

• Don't log unauthorized requests
• Log unauthorized requests without the request body

You can easily achieve the former setup by placing the LogbookFilter after your security filter. The latter is a little bit more sophisticated. You’ll need two LogbookFilter instances — one before your security filter, and one after it:

context.addFilter("SecureLogbookFilter", new SecureLogbookFilter(logbook))
    .addMappingForUrlPatterns(EnumSet.of(REQUEST, ASYNC), true, "/*");
context.addFilter("securityFilter", new SecurityFilter())
    .addMappingForUrlPatterns(EnumSet.of(REQUEST), true, "/*");
context.addFilter("LogbookFilter", new LogbookFilter(logbook))
    .addMappingForUrlPatterns(EnumSet.of(REQUEST, ASYNC), true, "/*");

The first logbook filter will log unauthorized requests only. The second filter will log authorized requests, as always.

HTTP Client

The logbook-httpclient module contains both an HttpRequestInterceptor and an HttpResponseInterceptor to use with the HttpClient:

CloseableHttpClient client = HttpClientBuilder.create()
        .addInterceptorFirst(new LogbookHttpRequestInterceptor(logbook))
        .addInterceptorFirst(new LogbookHttpResponseInterceptor())
        .build();

Since the LogbookHttpResponseInterceptor is incompatible with the HttpAsyncClient there is another way to log responses:

CloseableHttpAsyncClient client = HttpAsyncClientBuilder.create()
        .addInterceptorFirst(new LogbookHttpRequestInterceptor(logbook))
        .build();
        
// and then wrap your response consumer
client.execute(producer, new LogbookHttpAsyncResponseConsumer<>(consumer), callback)

JAX-RS

The logbook-jaxrs module contains:

A LogbookClientFilter to be used for applications making HTTP requests

client.register(new LogbookClientFilter(logbook));

A LogbookServerFilter for be used with HTTP servers

resourceConfig.register(new LogbookServerFilter(logbook));

Netty

The logbook-netty module contains:

A LogbookClientHandler to be used with an HttpClient:

HttpClient.create()
    .tcpConfiguration(tcpClient ->
        tcpClient.doOnConnected(connection ->
            connection.addHandlerLast(new LogbookClientHandler(logbook))))

A LogbookServerHandler for use used with an HttpServer:

HttpServer.create()
    .tcpConfiguration(tcpServer ->
        tcpServer.doOnConnection(connection ->
            connection.addHandlerLast(new LogbookServerHandler(logbook))))

Spring WebFlux

Users of Spring WebFlux can pick any of the following options:

• Programmatically create a NettyWebServer (passing an HttpServer)
• Register a custom NettyServerCustomizer
• Programmatically create a ReactorClientHttpConnector (passing an HttpClient)
• Register a custom WebClientCustomizer

Micronaut

Users of Micronaut can follow the official docs on how to integrate Logbook with Micronaut.

⚠️ Even though Quarkus and Vert.x use Netty under the hood, unfortunately neither of them allows accessing or customizing it (yet).

OkHttp v2.x

The logbook-okhttp2 module contains an Interceptor to use with version 2.x of the OkHttpClient:

OkHttpClient client = new OkHttpClient();
client.networkInterceptors().add(new LogbookInterceptor(logbook);

If you're expecting gzip-compressed responses you need to register our GzipInterceptor in addition. The transparent gzip support built into OkHttp will run after any network interceptor which forces logbook to log compressed binary responses.

OkHttpClient client = new OkHttpClient();
client.networkInterceptors().add(new LogbookInterceptor(logbook);
client.networkInterceptors().add(new GzipInterceptor());

OkHttp v3.x

The logbook-okhttp module contains an Interceptor to use with version 3.x of the OkHttpClient:

OkHttpClient client = new OkHttpClient.Builder()
        .addNetworkInterceptor(new LogbookInterceptor(logbook))
        .build();

If you're expecting gzip-compressed responses you need to register our GzipInterceptor in addition. The transparent gzip support built into OkHttp will run after any network interceptor which forces logbook to log compressed binary responses.

OkHttpClient client = new OkHttpClient.Builder()
        .addNetworkInterceptor(new LogbookInterceptor(logbook))
        .addNetworkInterceptor(new GzipInterceptor())
        .build();

Spring

The logbook-spring module contains a ClientHttpRequestInterceptor to use with RestTemplate:

    LogbookClientHttpRequestInterceptor interceptor = new LogbookClientHttpRequestInterceptor(logbook);
    RestTemplate restTemplate = new RestTemplate();
    restTemplate.getInterceptors().add(interceptor);

Spring Boot Starter

Logbook comes with a convenient auto configuration for Spring Boot users. It sets up all of the following parts automatically with sensible defaults:

• Servlet filter
• Second Servlet filter for unauthorized requests (if Spring Security is detected)
• Header-/Parameter-/Body-Filters
• HTTP-/JSON-style formatter
• Logging writer

Instead of declaring a dependency to logbook-core declare one to the Spring Boot Starter:

<dependency>
    <groupId>org.zalando</groupId>
    <artifactId>logbook-spring-boot-starter</artifactId>
    <version>${logbook.version}</version>
</dependency>

Every bean can be overridden and customized if needed, e.g. like this:

@Bean
public BodyFilter bodyFilter() {
    return merge(
            defaultValue(), 
            replaceJsonStringProperty(singleton("secret"), "XXX"));
}

Please refer to LogbookAutoConfiguration or the following table to see a list of possible integration points:

Multiple filters are merged into one.

Autoconfigured beans from logbook-spring

Some classes from logbook-spring are included in the auto configuration.

You can autowire LogbookClientHttpRequestInterceptor with code like:

private final RestTemplate restTemplate;
MyClient(RestTemplateBuilder builder, LogbookClientHttpRequestInterceptor interceptor){
  this.restTemplate = builder
    .additionalInterceptors(interceptor)
    .build();
}

Configuration

The following tables show the available configuration:

Example configuration

logbook:
  include:
    - /api/**
    - /actuator/**
  exclude:
    - /actuator/health
    - /api/admin/**
  filter.enabled: true
  secure-filter.enabled: true
  format.style: http
  strategy: body-only-if-status-at-least
  minimum-status: 400
  obfuscate:
    headers:
      - Authorization
      - X-Secret
    parameters:
      - access_token
      - password
  write:
    chunk-size: 1000

logstash-logback-encoder

For basic Logback configuraton

appender name="STDOUT" class="ch.qos.logback.core.ConsoleAppender">
   <encoder class="net.logstash.logback.encoder.LogstashEncoder"/>
/appender>

configure Logbook with a LogstashLogbackSink

HttpLogFormatter formatter = new JsonHttpLogFormatter();
LogstashLogbackSink sink = new LogstashLogbackSink(formatter);

for outputs like

{
  "@timestamp" : "2019-03-08T09:37:46.239+01:00",
  "@version" : "1",
  "message" : "GET http://localhost/test?limit=1",
  "logger_name" : "org.zalando.logbook.Logbook",
  "thread_name" : "main",
  "level" : "TRACE",
  "level_value" : 5000,
  "http" : {
     // logbook request/response contents
  }
}

Known Issues

1. The Logbook Servlet Filter interferes with downstream code using getWriter and/or getParameter*(). See Servlet for more details.
2. The Logbook Servlet Filter does NOT support ERROR dispatch. You're strongly encouraged to not use it to produce error responses.
3. The Logbook HTTP Client integration is handling gzip-compressed response entities incorrectly if the interceptor runs before a decompressing interceptor. Since logging compressed contents is not really helpful it's advised to register the logbook interceptor as the last interceptor in the chain.

Getting Help with Logbook

If you have questions, concerns, bug reports, etc., please file an issue in this repository's Issue Tracker.

Getting Involved/Contributing

To contribute, simply make a pull request and add a brief description (1-2 sentences) of your addition or change. For more details, check the contribution guidelines.

Alternatives

• Apache HttpClient Wire Logging
  • Client-side only
  • Apache HttpClient exclusive
  • Support for HTTP bodies
• Spring Boot Access Logging
  • Spring application only
  • Server-side only
  • Tomcat/Undertow/Jetty exclusive
  • No support for HTTP bodies
• Tomcat Request Dumper Filter
  • Server-side only
  • Tomcat exclusive
  • No support for HTTP bodies
• logback-access
  • Server-side only
  • Any servlet container
  • Support for HTTP bodies

Credits and References

Grand Turk, a replica of a three-masted 6th rate frigate from Nelson's days - logbook and charts by JoJan is licensed under a Creative Commons (Attribution-Share Alike 3.0 Unported).