Tomcat部署Spark Java遗留应用失败：SparkFilter问题排查咨询

Let’s break down your problem step by step—since you’ve already confirmed the Spark Java dependency is in your pom.xml, here are the key areas to check next:

1. Verify Dependency Scope & War Package Contents #

First, double-check how your Spark Java dependency is configured in pom.xml:

• Make sure you haven’t set the <scope> to provided unless you’ve manually added the Spark JARs to Tomcat’s global lib directory. If the scope is provided, the Spark classes (including spark.servlet.SparkFilter) won’t be packaged into your WAR file, leading to a ClassNotFoundException.
• Example of a correct dependency entry:

  <dependency>
      <groupId>com.sparkjava</groupId>
      <artifactId>spark-core</artifactId>
      <version>2.9.4</version> <!-- Use your actual version -->
  </dependency>
  

• Extract your WAR file and check WEB-INF/lib—you should see spark-core-[version].jar present. If it’s missing, your build process isn’t including the dependency correctly (check for Maven profile issues or excluded dependencies).

2. Fix the Root Cause: NullPointerException in Your Custom Spark Class #

The stack trace points to a NullPointerException at com.example.resource.Spark.init(Spark.java:121)—this is your application code, not the Spark framework. Here’s what to check:

• Look at line 121 of your Spark class: Are you referencing an object that hasn’t been initialized yet? For example, if you’re using properties loaded by your PropertiesLoader listener, ensure the listener has fully completed initialization before your Spark app tries to access those properties.
• Confirm your com.example.resource.Spark class correctly implements the spark.servlet.SparkApplication interface. SparkFilter expects this interface to call the init() method—if you’re using a custom static init method instead, you might run into initialization order issues.
• Add debug logs or attach a debugger to Tomcat to inspect which variable is null at line 121. This will directly point you to the problem.

3. Validate Web.xml Configuration #

• Double-check the applicationClass parameter value: com.example.resource.Spark must be the fully qualified name of your Spark application class. Typos here can lead to class loading issues or initialization failures.
• Ensure the filter-class entry (spark.servlet.SparkFilter) is correct—this is the standard class name for Spark’s Tomcat filter, so it should match exactly.
• Confirm the listener (com.example.utilities.PropertiesLoader) is properly configured and doesn’t throw any silent exceptions during initialization. If the listener fails to load properties, your Spark app might try to access null values later.

4. Check Tomcat Class Loading & Version Compatibility #

• Avoid placing Spark JARs in both your WAR’s WEB-INF/lib and Tomcat’s global lib directory—this can cause class loading conflicts (e.g., duplicate class definitions). Pick one location and stick with it.
• Verify compatibility between your Spark Java version and Tomcat version: For example, Spark Java 2.x requires Servlet 3.0+ (compatible with Tomcat 7+), while older Spark versions might work with Servlet 2.5 (Tomcat 6). If you’re using a newer Tomcat version, make sure your Spark Java version supports the corresponding Servlet spec.

5. Look for Earlier Errors in Catalina.out #

The NPE you’re seeing might be a secondary issue—check the full catalina.out log for earlier ClassNotFoundException entries related to SparkFilter. If such errors exist, they indicate the Spark JAR isn’t being loaded correctly, which you can fix by addressing step 1 above.

内容的提问来源于stack exchange，提问作者Preethi Jha