[GEODE-6630] RedundancyLogger might throw NPE

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.9.0
• Fix Version/s: 1.9.0
• Labels: SmallFeature
• Assignee: Eric Shu
• Reporter: Xiaojian Zhou
• Created: 10/Apr/19 20:44
• Updated: 28/May/19 19:55
• Resolved: 18/Apr/19 23:56

Description

I noticed in some tests with following error logged in debug level:

[debug 2019/04/10 11:25:52.904 PDT <RedundancyLogger for region TestRegion> tid=0x55] Unexpected exception in PR redundancy recovery
java.lang.NullPointerException
        at org.apache.geode.internal.cache.partitioned.RedundancyLogger.run2(RedundancyLogger.java:154)
        at org.apache.geode.internal.cache.partitioned.RecoveryRunnable.run(RecoveryRunnable.java:59)
        at java.lang.Thread.run(Thread.java:748)

Then I looked at the code that it's using:
(this.allBucketsRecoveredFromDisk.getCount() > 0) in more than one places.

The allBucketsRecoveredFromDisk is referecing to the object in PRHARedundancyProvider. 
But in PRHARedundancyProvider, it is using:
  public boolean isPersistentRecoveryComplete() {
    if (!ColocationHelper.checkMembersColocation(this.prRegion, this.prRegion.getMyId())) {
      return false;
    }

    if (allBucketsRecoveredFromDisk != null && allBucketsRecoveredFromDisk.getCount() > 0) {
      return false;
    }

So allBucketsRecoveredFromDisk could be null. We should alway use 
"(allBucketsRecoveredFromDisk != null && allBucketsRecoveredFromDisk.getCount() > 0) " instead of "(this.allBucketsRecoveredFromDisk.getCount() > 0)"

Issue Links

• links to: GitHub Pull Request #3477