[hbase] branch branch-2.2 updated: HBASE-24897 RegionReplicaFlushHandler should handle NoServerForRegionException to avoid aborting RegionServer (#2271)

[zg...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

zghao pushed a commit to branch branch-2.2
in repository https://gitbox.apache.org/repos/asf/hbase.git


The following commit(s) were added to refs/heads/branch-2.2 by this push:
     new 0bbb830  HBASE-24897 RegionReplicaFlushHandler should handle NoServerForRegionException to avoid aborting RegionServer (#2271)
0bbb830 is described below

commit 0bbb83060646506250228940cb710bc834745114
Author: Guanghao Zhang <zg...@apache.org>
AuthorDate: Wed Aug 19 18:27:12 2020 +0800

    HBASE-24897 RegionReplicaFlushHandler should handle NoServerForRegionException to avoid aborting RegionServer (#2271)
    
    Signed-off-by: Duo Zhang <zh...@apache.org>
---
 .../regionserver/handler/RegionReplicaFlushHandler.java   | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/RegionReplicaFlushHandler.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/RegionReplicaFlushHandler.java
index 81b6d7e..71c112d 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/RegionReplicaFlushHandler.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/handler/RegionReplicaFlushHandler.java
@@ -25,6 +25,8 @@ import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.hbase.HConstants;
 import org.apache.hadoop.hbase.Server;
 import org.apache.hadoop.hbase.TableNotFoundException;
+import org.apache.hadoop.hbase.client.NoServerForRegionException;
+import org.apache.hadoop.hbase.client.RegionInfo;
 import org.apache.yetus.audience.InterfaceAudience;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -115,11 +117,12 @@ public class RegionReplicaFlushHandler extends EventHandler {
         .getRegionInfoForDefaultReplica(region.getRegionInfo()).getEncodedName() + " of region "
        + region.getRegionInfo().getEncodedName() + " to trigger a flush");
     }
+    RegionInfo primaryRegion =
+        RegionReplicaUtil.getRegionInfoForDefaultReplica(region.getRegionInfo());
     while (!region.isClosing() && !region.isClosed()
         && !server.isAborted() && !server.isStopped()) {
-      FlushRegionCallable flushCallable = new FlushRegionCallable(
-        connection, rpcControllerFactory,
-        RegionReplicaUtil.getRegionInfoForDefaultReplica(region.getRegionInfo()), true);
+      FlushRegionCallable flushCallable =
+          new FlushRegionCallable(connection, rpcControllerFactory, primaryRegion, true);
 
       // TODO: flushRegion() is a blocking call waiting for the flush to complete. Ideally we
       // do not have to wait for the whole flush here, just initiate it.
@@ -132,6 +135,12 @@ public class RegionReplicaFlushHandler extends EventHandler {
             || connection.isTableDisabled(region.getRegionInfo().getTable())) {
           return;
         }
+        // This may happen when create a table with region replica. And the CreateTableProcedure
+        // assign secondary regions firstly, then assign primary region. See HBASE-24897
+        if (ex instanceof NoServerForRegionException) {
+          LOG.warn("Primary region {} is not online, retry", primaryRegion, ex);
+          continue;
+        }
         throw ex;
       }