src/main/java/com/kamco/cd/training/train/service/DockerTrainService.java

@@ -62,10 +62,36 @@ public class DockerTrainService {
     ProcessBuilder pb = new ProcessBuilder(cmd);
     pb.redirectErrorStream(true);
 
-    Process p = pb.start();
-
     // 로그는 별도 스레드에서 읽기 (메인 스레드가 readLine에 안 걸리게)
     StringBuilder logBuilder = new StringBuilder();
+    Process p = pb.start();
+
+    log.info("[TRAIN-BOOT] docker run started. container={}", containerName);
+
+    try {
+      log.info("[TRAIN-BOOT] pid={}", p.pid()); // Java 9+
+    } catch (Throwable ignore) {
+    }
+
+    try {
+      // 바로 죽었는지 100ms만 체크
+      if (p.waitFor(100, TimeUnit.MILLISECONDS)) {
+        int exit = p.exitValue();
+        String earlyLogs;
+        synchronized (logBuilder) {
+          earlyLogs = logBuilder.toString();
+        }
+        log.error(
+            "[TRAIN-BOOT] docker run exited immediately. container={} exit={}",
+            containerName,
+            exit);
+        log.error("[TRAIN-BOOT] early logs:\n{}", earlyLogs);
+      } else {
+        log.info("[TRAIN-BOOT] docker run is still running. container={}", containerName);
+      }
+    } catch (Exception e) {
+      log.warn("[TRAIN-BOOT] early-exit check failed: {}", e.toString(), e);
+    }
 
     Pattern epochPattern = Pattern.compile("Epoch\\(train\\)\\s+\\[(\\d+)\\]\\[(\\d+)/(\\d+)\\]");
 
@@ -176,7 +202,7 @@ public class DockerTrainService {
 
     List<String> c = new ArrayList<>();
 
-    c.add("docker");
+    c.add("/usr/bin/docker");
     c.add("run");
 
     // 컨테이너 이름 지정
@@ -386,7 +412,7 @@ public class DockerTrainService {
 
     List<String> c = new ArrayList<>();
 
-    c.add("docker");
+    c.add("/usr/bin/docker");
     c.add("run");
     c.add("--name");
     c.add(containerName);