fix when the job fails and doesn't kick the node (Azure#3186)

* fix when the job fails and doesn't kick the node

* reformat

Author: cassieesvelt

best-practices/largescale-deep-learning/Debugging/Compute/run_nhc.py

@@ -100,6 +100,7 @@ def parse_output(output_file):
             output, error = process.communicate()
             print(output)
             print(error)
+        if full_errors:
             raise Exception(
                 "Failures were found while running the node health checks. Please see the std_log_process.txt files under the 'outputs and logs' tab of the job for more information."
                 + full_errors