Fix exitCode check (#190)

The exitCode check works differently then described in the API: https://github.com/kubeflow/common/blob/master/pkg/apis/common/v1/types.go\#L163-L168

In addition the document exitCodes for Docker will not appear in a Kubernetes pod status (anymore?)

pkg/util/train/train_util.go

@@ -16,38 +16,5 @@
 package train
 
 func IsRetryableExitCode(exitCode int32) bool {
-	if exitCode == 1 || exitCode == 2 || exitCode == 126 ||
-		exitCode == 127 || exitCode == 128 || exitCode == 139 {
-		// Refers to http://tldp.org/LDP/abs/html/exitcodes.html, we identify the following exit codes
-		// as permanent errors:
-		//   1: General errors
-		//   2: Misuse of shell builtins
-		//   126: Command invoked cannot execute
-		//   127: Command not found
-		//   128: Invalid argument to exit
-		//   139(128+11): terminated by SIGSEGV(Invalid memory reference)
-		return false
-	}
-
-	if exitCode == 130 || exitCode == 137 || exitCode == 143 {
-		// We think it's retryable error if the container exits due to the following sys signals
-		// that are usually caused by transient issues(e.g. VM was rescheduled):
-		//   130(128+2): Container terminated by Control-C
-		//   137(128+9): Container received a SIGKILL
-		//   143(128+15): Container received a SIGTERM
-		// The exit code of container will be 128 + n for fatal error signals.
-		// More info can be found in:
-		//   http://tldp.org/LDP/abs/html/exitcodes.html,
-		//   https://stackoverflow.com/questions/31297616/what-is-the-authoritative-list-of-docker-run-exit-codes
-		return true
-	}
-
-	if exitCode == 138 {
-		// We allow users to specify exit code for the cases that they think should retry.
-		// We decide to take the exit code of SIGUSR1(138 = 128 + 10) for user defined retryable error.
-		return true
-	}
-
-	// We make no guarantee for other exit status. Currently handling them same as permanent errors.
-	return false
+	return exitCode >= 128
 }