[HBASE-23376] NPE happens while replica region is moving - ASF JIRA

XML

Word

Printable

JSON

Details

Type: Bug
Status: Resolved
Priority: Minor
Resolution: Fixed
Affects Version/s: None
Fix Version/s: 3.0.0-alpha-1, 2.3.0, 2.2.3, 2.1.9
Component/s: read replicas
Labels:
None

Hadoop Flags:

Reviewed

Description

The following code is from AsyncNonMetaRegionLocator#addToCache

private RegionLocations addToCache(TableCache tableCache, RegionLocations locs) {
  LOG.trace("Try adding {} to cache", locs);
  byte[] startKey = locs.getDefaultRegionLocation().getRegion().getStartKey();
  ...
}

we will get a NPE if the locs is without the default region.

The following code is from AsyncRegionLocatorHelper#updateCachedLocationOnError

...
if (cause instanceof RegionMovedException) {
  RegionMovedException rme = (RegionMovedException) cause;
  HRegionLocation newLoc =
    new HRegionLocation(loc.getRegion(), rme.getServerName(), rme.getLocationSeqNum());
  LOG.debug("Try updating {} with the new location {} constructed by {}", loc, newLoc,
    rme.toString());
  addToCache.accept(newLoc);
...

If the replica region is moving, we will get a RegionMovedException and add the HRegionLocation of replica region to cache. And finally NPE happens.

java.lang.NullPointerExceptionjava.lang.NullPointerException at org.apache.hadoop.hbase.client.AsyncNonMetaRegionLocator.addToCache(AsyncNonMetaRegionLocator.java:240) at org.apache.hadoop.hbase.client.AsyncNonMetaRegionLocator.addLocationToCache(AsyncNonMetaRegionLocator.java:596) at org.apache.hadoop.hbase.client.AsyncRegionLocatorHelper.updateCachedLocationOnError(AsyncRegionLocatorHelper.java:80) at org.apache.hadoop.hbase.client.AsyncNonMetaRegionLocator.updateCachedLocationOnError(AsyncNonMetaRegionLocator.java:610) at org.apache.hadoop.hbase.client.AsyncRegionLocator.updateCachedLocationOnError(AsyncRegionLocator.java:153)