[HADOOP-135] Potential deadlock in JobTracker.

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.1.0
• Fix Version/s: 0.4.0
• Assignee: Owen O'Malley
• Reporter: Konstantin Shvachko
• Created: 14/Apr/06 07:28
• Updated: 08/Jul/09 16:51
• Resolved: 27/Jun/06 00:35

Description

org.apache.hadoop.mapred.JobTracker.RetireJobs.run()
locks resources in this order
synchronized (jobs) {
synchronized (jobInitQueue) {
synchronized (jobsByArrival) {

org.apache.hadoop.mapred.JobTracker.submitJob()
locks resources in a different order
synchronized (jobs) {
synchronized (jobsByArrival) {
synchronized (jobInitQueue) {

This potentially can lead to a deadlock.
Unless there is common lock on top of it in which case these
three locks are redundant.