[hbase] branch branch-2 updated: HBASE-25837 TestRollingRestart is flaky (#3220)

[niuyulin]

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

niuyulin pushed a commit to branch branch-2
in repository https://gitbox.apache.org/repos/asf/hbase.git


The following commit(s) were added to refs/heads/branch-2 by this push:
     new 5e65da6  HBASE-25837 TestRollingRestart is flaky (#3220)
5e65da6 is described below

commit 5e65da64e229b737b71541c8bbc4a5514760d525
Author: niuyulin <yulin.niu.2...@gmail.com>
AuthorDate: Fri May 7 18:58:45 2021 +0800

    HBASE-25837 TestRollingRestart is flaky (#3220)
    
    Signed-off-by: Duo Zhang <zhang...@apache.org>
---
 .../hadoop/hbase/master/TestRollingRestart.java       | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git 
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestRollingRestart.java
 
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestRollingRestart.java
index 7530121..3ac054c 100644
--- 
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestRollingRestart.java
+++ 
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestRollingRestart.java
@@ -38,6 +38,7 @@ import org.apache.hadoop.hbase.TableName;
 import org.apache.hadoop.hbase.client.RegionInfo;
 import org.apache.hadoop.hbase.client.RegionLocator;
 import org.apache.hadoop.hbase.client.Table;
+import org.apache.hadoop.hbase.master.procedure.ServerCrashProcedure;
 import org.apache.hadoop.hbase.testclassification.LargeTests;
 import org.apache.hadoop.hbase.testclassification.MasterTests;
 import org.apache.hadoop.hbase.util.Bytes;
@@ -68,6 +69,7 @@ public class TestRollingRestart {
 
   private static final Logger LOG = 
LoggerFactory.getLogger(TestRollingRestart.class);
 
+  private static HBaseTestingUtility TEST_UTIL;
   @Rule
   public TestName name = new TestName();
 
@@ -89,7 +91,7 @@ public class TestRollingRestart {
     Configuration conf = HBaseConfiguration.create();
     conf.setBoolean(HConstants.HBASE_SPLIT_WAL_COORDINATED_BY_ZK,
         splitWALCoordinatedByZK);
-    HBaseTestingUtility TEST_UTIL = new HBaseTestingUtility(conf);
+    TEST_UTIL = new HBaseTestingUtility(conf);
     StartMiniClusterOption option = StartMiniClusterOption.builder()
         
.numMasters(NUM_MASTERS).numRegionServers(NUM_RS).numDataNodes(NUM_RS).build();
     TEST_UTIL.startMiniCluster(option);
@@ -218,8 +220,18 @@ public class TestRollingRestart {
     TEST_UTIL.shutdownMiniCluster();
   }
 
+  /**
+   * Checks if the SCP of specific dead server has been executed.
+   * @return true if the SCP of specific serverName has been executed, false 
if not
+   */
+  private boolean isDeadServerSCPExecuted(ServerName serverName) throws 
IOException {
+    return TEST_UTIL.getMiniHBaseCluster().getMaster().getProcedures().stream()
+        .anyMatch(p -> p instanceof ServerCrashProcedure
+            && ((ServerCrashProcedure) p).getServerName().equals(serverName));
+  }
+
   private void waitForRSShutdownToStartAndFinish(MasterThread activeMaster,
-      ServerName serverName) throws InterruptedException {
+      ServerName serverName) throws InterruptedException, IOException {
     ServerManager sm = activeMaster.getMaster().getServerManager();
     // First wait for it to be in dead list
     while (!sm.getDeadServers().isDeadServer(serverName)) {
@@ -228,6 +240,9 @@ public class TestRollingRestart {
     }
     log("Server [" + serverName + "] marked as dead, waiting for it to " +
         "finish dead processing");
+
+    TEST_UTIL.waitFor(60000, () -> isDeadServerSCPExecuted(serverName));
+
     while (sm.areDeadServersInProgress()) {
       log("Server [" + serverName + "] still being processed, waiting");
       Thread.sleep(100);