[LUCENE-6670] OOM error in merge causes deadlock

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 5.2.1
• Fix Version/s: 5.3, 6.0
• Lucene Fields: New
• Assignee: Unassigned
• Reporter: Jeffrey Morlan
• Created: 09/Jul/15 17:42
• Updated: 28/Aug/22 14:38
• Resolved: 14/Jul/15 09:44

Description

Using the ConcurrentMergeScheduler, when merging catches an OutOfMemoryError and tries to rollback, the merge thread ends up calling Thread.join on itself, which will never complete.

Stack trace (line numbers are for Lucene 5.2.1):

"Lucene Merge Thread #1" #356502 daemon prio=5 os_prio=0 tid=0x00007f3f78622000 nid=0x11b5 in Object.wait() [0x00007f3e90528000]
   java.lang.Thread.State: WAITING (on object monitor)
        at java.lang.Object.wait(Native Method)
        at java.lang.Thread.join(Thread.java:1245)
        - locked <0x00000001965aff78> (a org.apache.lucene.index.ConcurrentMergeScheduler$MergeThread)
        at java.lang.Thread.join(Thread.java:1319)
        at org.apache.lucene.index.ConcurrentMergeScheduler.sync(ConcurrentMergeScheduler.java:420)
        at org.apache.lucene.index.ConcurrentMergeScheduler.close(ConcurrentMergeScheduler.java:401)
        at org.apache.lucene.index.IndexWriter.rollbackInternal(IndexWriter.java:1948)
        at org.apache.lucene.index.IndexWriter.rollback(IndexWriter.java:1921)
        - locked <0x00000001965b0100> (a java.lang.Object)
        at org.apache.lucene.index.IndexWriter.tragicEvent(IndexWriter.java:4439)
        - locked <0x00000001965b0100> (a java.lang.Object)
        at org.apache.lucene.index.IndexWriter.merge(IndexWriter.java:3571)
        at org.apache.lucene.index.ConcurrentMergeScheduler.doMerge(ConcurrentMergeScheduler.java:581)
        at org.apache.lucene.index.ConcurrentMergeScheduler$MergeThread.run(ConcurrentMergeScheduler.java:619)

Issue Links

• is duplicated by: LUCENE-6579 Unexpected merge exceptions should be tragic to IndexWriter