task-152: 清理失败吞异常记指标（aiimage.temp-cleanup.failed 计数、日志+指标均不阻断主流程、按路径隔离、无重试循环）+ 8 条测试

2026-09-02 06:29:59 +08:00
parent f0e12b9b1b
commit 3b34ec3bef
2 changed files with 144 additions and 1 deletions
@@ -2,8 +2,10 @@ package com.nanri.aiimage.modules.file.service;

 import cn.hutool.core.io.FileUtil;
 import com.nanri.aiimage.config.StorageProperties;
+import io.micrometer.core.instrument.MeterRegistry;
 import lombok.RequiredArgsConstructor;
 import lombok.extern.slf4j.Slf4j;
+import org.springframework.beans.factory.annotation.Autowired;
 import org.springframework.scheduling.annotation.Scheduled;
 import org.springframework.stereotype.Service;

@@ -20,6 +22,7 @@ public class LocalTempCleanupService {

    private static final Pattern ROOT_TEMP_FILE_PATTERN = Pattern.compile("^[a-fA-F0-9]{32}(\\.[^.]+)?$");
    private static final String TRANSIENT_PAYLOAD_DIR_NAME = "transient-payload";
+    private static final String CLEANUP_FAILED_METRIC = "aiimage.temp-cleanup.failed";
    private static final List<String> RESULT_DIR_NAMES = List.of(
            "dedupe-result",
            "convert-result",
@@ -32,6 +35,10 @@ public class LocalTempCleanupService {

    private final StorageProperties storageProperties;

+    /** 指标注册表（可选注入：无注册表时仅记日志，不影响清理）。 */
+    @Autowired(required = false)
+    private MeterRegistry meterRegistry;
+
    @Scheduled(cron = "${aiimage.storage.cleanup-cron:0 0 */6 * * *}")
    public void cleanupLocalTempDir() {
        if (!storageProperties.isCleanupEnabled()) {
@@ -78,7 +85,7 @@ public class LocalTempCleanupService {
                    deleteEmptyDirectories(child, tempDir);
                }
            } catch (Exception ex) {
-                log.warn("local temp cleanup failed: path={}", child.getAbsolutePath(), ex);
+                handleCleanupFailure(child, ex);
            }
        }

@@ -125,6 +132,23 @@ public class LocalTempCleanupService {
        return ROOT_TEMP_FILE_PATTERN.matcher(file.getName()).matches();
    }

+    /**
+     * 清理失败处理：记录日志与失败指标，不抛异常、不重试（下一轮定时清理自然重试）。
+     * 单个条目失败不影响其余条目的清理。
+     */
+    void handleCleanupFailure(File child, Exception ex) {
+        log.warn("local temp cleanup failed: path={}", child == null ? null : child.getAbsolutePath(), ex);
+        try {
+            if (meterRegistry != null && child != null) {
+                meterRegistry.counter(CLEANUP_FAILED_METRIC, "path", child.getName()).increment();
+            }
+        } catch (Exception metricEx) {
+            // 指标记录失败也不阻断清理主流程
+            log.warn("local temp cleanup metric record failed path={}",
+                    child == null ? null : child.getName(), metricEx);
+        }
+    }
+
    private boolean isExpired(File file, Instant expireBefore) {
        // 以 mtime 近似最后访问（TempFileMetadata 回退语义，与现状一致）
        return TempFileMetadata.lastAccessTime(file).isBefore(expireBefore);
@@ -0,0 +1,119 @@
+package com.nanri.aiimage.modules.file.service;
+
+import com.nanri.aiimage.config.StorageProperties;
+import io.micrometer.core.instrument.Counter;
+import io.micrometer.core.instrument.MeterRegistry;
+import org.junit.jupiter.api.BeforeEach;
+import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.extension.ExtendWith;
+import org.mockito.Mock;
+import org.mockito.junit.jupiter.MockitoExtension;
+import org.springframework.test.util.ReflectionTestUtils;
+
+import java.io.File;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.mockito.ArgumentMatchers.anyString;
+import static org.mockito.ArgumentMatchers.eq;
+import static org.mockito.Mockito.mock;
+import static org.mockito.Mockito.times;
+import static org.mockito.Mockito.verify;
+import static org.mockito.Mockito.when;
+
+/**
+ * task-152：清理失败吞异常记指标契约（plan 09）。
+ * 清理失败：记录日志+指标（清理失败计数）、不抛异常、不自动重试成循环；
+ * 单个条目失败不影响其他条目（隔离）。
+ */
+@ExtendWith(MockitoExtension.class)
+class CleanupFailureMetricTest {
+
+    @Mock private StorageProperties storageProperties;
+    @Mock private MeterRegistry meterRegistry;
+
+    private LocalTempCleanupService service;
+
+    @BeforeEach
+    void setUp() {
+        service = new LocalTempCleanupService(storageProperties);
+        ReflectionTestUtils.setField(service, "meterRegistry", meterRegistry);
+    }
+
+    @Test
+    void failureIsLoggedWithoutThrowing() {
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom"));
+        service.handleCleanupFailure(new File("target/tmp/b.tmp"), null);
+    }
+
+    @Test
+    void failureRecordsMetric() {
+        Counter counter = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), anyString(), anyString())).thenReturn(counter);
+
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom"));
+
+        verify(meterRegistry).counter(eq("aiimage.temp-cleanup.failed"), eq("path"), eq("a.tmp"));
+        verify(counter).increment();
+    }
+
+    @Test
+    void failureDoesNotThrowToCaller() {
+        Counter counter = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), anyString(), anyString())).thenThrow(
+                new RuntimeException("metrics down"));
+
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom"));
+    }
+
+    @Test
+    void failureMetricNotRepeatedInLoop() {
+        Counter counter = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), anyString(), anyString())).thenReturn(counter);
+
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom"));
+
+        verify(counter, times(1)).increment();
+    }
+
+    @Test
+    void nextRoundCanRetryIndependently() {
+        Counter counter = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), anyString(), anyString())).thenReturn(counter);
+
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom"));
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("boom again"));
+
+        verify(counter, times(2)).increment();
+    }
+
+    @Test
+    void partialFailureIsIsolatedPerPath() {
+        Counter counterA = mock(Counter.class);
+        Counter counterB = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), eq("path"), eq("a.tmp"))).thenReturn(counterA);
+        when(meterRegistry.counter(anyString(), eq("path"), eq("b.tmp"))).thenReturn(counterB);
+
+        service.handleCleanupFailure(new File("target/tmp/a.tmp"), new RuntimeException("x"));
+        service.handleCleanupFailure(new File("target/tmp/b.tmp"), new RuntimeException("y"));
+        service.handleCleanupFailure(new File("target/tmp/b.tmp"), new RuntimeException("y2"));
+
+        verify(counterA, times(1)).increment();
+        verify(counterB, times(2)).increment();
+    }
+
+    @Test
+    void metricTaggedByFileName() {
+        Counter counter = mock(Counter.class);
+        when(meterRegistry.counter(anyString(), anyString(), anyString())).thenReturn(counter);
+
+        service.handleCleanupFailure(new File("target/tmp/result/x.xlsx"), new RuntimeException("boom"));
+
+        verify(meterRegistry).counter(eq("aiimage.temp-cleanup.failed"), eq("path"), eq("x.xlsx"));
+    }
+
+    @Test
+    void nullChildIsSafe() {
+        service.handleCleanupFailure(null, new RuntimeException("boom"));
+        service.handleCleanupFailure(null, null);
+    }
+}