Michael Perez created HBASE-29221:
-------------------------------------
Summary: ClientSideRegionScanner fails with NPE while trying to
close mslab
Key: HBASE-29221
URL: https://issues.apache.org/jira/browse/HBASE-29221
Project: HBase
Issue Type: Bug
Components: Client, Scanners
Reporter: Michael Perez
{code:java}
java.io.IOException: java.lang.NullPointerException: Cannot invoke
"org.apache.hadoop.hbase.regionserver.ChunkCreator.putbackChunks(java.util.Set)"
because "this.chunkCreator" is null
at org.apache.hadoop.hbase.regionserver.HRegion.doClose(HRegion.java:1877)
at org.apache.hadoop.hbase.regionserver.HRegion.close(HRegion.java:1626)
at org.apache.hadoop.hbase.regionserver.HRegion.close(HRegion.java:1581)
at org.apache.hadoop.hbase.regionserver.HRegion.close(HRegion.java:1564)
at
org.apache.hadoop.hbase.client.ClientSideRegionScanner.close(ClientSideRegionScanner.java:133)
at{code}
The ChunkCreator for the MemStoreLab is only initialized for [real HBase region
servers|https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/HBaseServerBase.java#L410-L431].
Instances of the ClientSideRegionScanner will never initialize the memstore
which causes NPE at close time.
This issue was caused by the changes in HBASE-28401
This same issue surfaced in HBASE-28743. The fix for that issue was to disable
the MemStoreLab for the TableSnapshotScanner ([PR
master|https://github.com/apache/hbase/pull/6101/files], [PR branch
2|https://github.com/apache/hbase/pull/6110]). The issue still persists in
cases where the ClientSideRegionScanner is used in any context other than as
part of the TableSnapshotScanner.
The proposed solution for this issue is to update [the
initialization|https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/client/ClientSideRegionScanner.java#L70]
of the ClientSideRegionScanner to disable the MemStoreLab.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
