Jenkins并行测试阶段报目录不存在错误的解决与重试方案咨询

Let's break this down step by step—your intermittent directory not found error for target/test-classes is almost certainly a race condition from parallel stages sharing the same workspace. Here's how to fix it, plus whether retry logic helps and how to implement it:

First: Root Cause & Core Fixes #

When you run stages in parallel on the same Jenkins node, all tasks are writing to and modifying the same target directory in the shared workspace. This causes occasional file system conflicts (like one task deleting the directory while another tries to write to it). Here are the most effective fixes:

1. Use Isolated Workspaces for Each Parallel Task #

Give every test stage its own subdirectory to avoid overlapping file operations. Modify your stageConstructor to wrap the test execution in a ws (workspace) step that targets a unique path per test:

def stageConstructor(String test, String testDesc, String nameSpace) {
    return {
        stage(testDesc) {
            // Create a unique workspace for each test to prevent conflicts
            ws("${env.WORKSPACE}/isolated-tests/${test}") {
                checkout scm // Check out code fresh in the isolated space
                echo "Executing ${testDesc} in namespace ${nameSpace}"
                // Run Maven with clean to ensure a fresh target directory
                sh "mvn -Dnamespace=${nameSpace} clean testCompile test -Dtest=${test}"
            }
        }
    }
}

This ensures each test has its own target directory, eliminating cross-task interference.

2. Explicitly Clean Before Compiling #

Add clean to your Maven command for every test. This guarantees you start with a fresh target directory each time, avoiding leftover files or directory states from previous runs that might cause conflicts.

3. Check for Node-Level Issues #

Occasionally, intermittent errors can stem from:

• Disk IO bottlenecks on the docker_test node (too many parallel tasks overwhelming the filesystem)
• Permissions issues for the Jenkins user on the workspace directory
• Temporary disk space shortages

Quick checks: Monitor the node's disk usage during runs, verify Jenkins has read/write access to /var/jenkins_home/workspace/test_master, and consider limiting the number of parallel tasks if the node is resource-constrained.

Does Retry Logic Help? #

Yes—retry logic is a great safety net for intermittent glitches (even after fixing the root cause). It can handle one-off file system race conditions, transient disk errors, or other temporary failures that slip through.

How to Implement Retry in Your Pipeline #

You can use Jenkins built-in retry step for simple cases, or a custom loop with delays for more control:

Option 1: Simple Built-in Retry #

Wrap your Maven command in a retry block to automatically retry on failure:

retry(3) { // Retry up to 3 times
    sh "mvn -Dnamespace=${nameSpace} clean testCompile test -Dtest=${test}"
}

Option 2: Custom Retry with Delay #

For more control (like adding a wait between retries to let the filesystem settle), use a loop:

int maxRetries = 3
int attempt = 0
boolean testPassed = false

while (!testPassed && attempt < maxRetries) {
    attempt++
    try {
        sh "mvn -Dnamespace=${nameSpace} clean testCompile test -Dtest=${test}"
        testPassed = true
        echo "${testDesc} passed on attempt ${attempt}"
    } catch (Exception e) {
        echo "${testDesc} failed on attempt ${attempt}: ${e.getMessage()}"
        if (attempt == maxRetries) {
            throw e // Re-throw after final attempt to mark the stage as failed
        }
        sleep(time: 20, unit: 'SECONDS') // Wait 20 seconds before retrying
    }
}

Bonus: Fix the Uninitialized listTests Variable #

I noticed in your pipeline code you're adding items to listTests but never initialize it. Add this line before your conditional blocks to avoid a NullPointerException:

List<String> listTests = []

Final Recommendation #

Start by implementing isolated workspaces (the core fix for race conditions), then add retry logic as a safety net. This combination will eliminate the majority of your intermittent errors while handling any remaining one-off issues.

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