Find the error and file that caused an EMR job to shut down

error-and-split.sh

#!/bin/bash

# Tested with EMR AMI 3.9.0, but most likely does not work except
# in fortuitous constellations of errors

failed_task_id=$(zgrep -F 'Task failed' steps/*/syslog.* | head -n 1 | perl -p -e 's/^.*(task_\S+).*$/$1/')
failed_task_log=$(zgrep -l -R -F $failed_task_id task-attempts | tail -n 1)
failed_attempt_id=$(zgrep -F FATAL $failed_task_log | head -n 1 | perl -p -e 's/^.*(attempt_\S+).*$/$1/')
failed_attempt_log=$(zgrep -l -R -F $failed_attempt_id task-attempts | tail -n 1)

zgrep -F 'Processing split' $failed_attempt_log
zgrep -A 10 -F 'Exception running child' $failed_attempt_log