You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@lucene.apache.org by sh...@apache.org on 2016/11/05 07:16:45 UTC

lucene-solr:master: SOLR-9716: RecoveryStrategy sends prep recovery command without setting read time out which can cause replica recovery to hang indefinitely on network partitions

Repository: lucene-solr
Updated Branches:
  refs/heads/master 358bdd490 -> 1f1990d8b


SOLR-9716: RecoveryStrategy sends prep recovery command without setting read time out which can cause replica recovery to hang indefinitely on network partitions


Project: http://git-wip-us.apache.org/repos/asf/lucene-solr/repo
Commit: http://git-wip-us.apache.org/repos/asf/lucene-solr/commit/1f1990d8
Tree: http://git-wip-us.apache.org/repos/asf/lucene-solr/tree/1f1990d8
Diff: http://git-wip-us.apache.org/repos/asf/lucene-solr/diff/1f1990d8

Branch: refs/heads/master
Commit: 1f1990d8be9fbbe0d95a10f3be1dffccec969a32
Parents: 358bdd4
Author: Shalin Shekhar Mangar <sh...@apache.org>
Authored: Sat Nov 5 12:46:42 2016 +0530
Committer: Shalin Shekhar Mangar <sh...@apache.org>
Committed: Sat Nov 5 12:46:42 2016 +0530

----------------------------------------------------------------------
 solr/CHANGES.txt                                |  3 ++
 .../org/apache/solr/cloud/RecoveryStrategy.java | 49 ++++++++++++++------
 .../solr/handler/admin/PrepRecoveryOp.java      |  3 ++
 .../org/apache/solr/util/TestInjection.java     | 32 +++++++++++++
 .../apache/solr/cloud/TestCloudRecovery.java    |  8 ++++
 5 files changed, 81 insertions(+), 14 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/1f1990d8/solr/CHANGES.txt
----------------------------------------------------------------------
diff --git a/solr/CHANGES.txt b/solr/CHANGES.txt
index 0411092..af97609 100644
--- a/solr/CHANGES.txt
+++ b/solr/CHANGES.txt
@@ -114,6 +114,9 @@ Bug Fixes
 * SOLR-9360: Solr script not properly checking SOLR_PID
   (Alessandro Benedetti via Erick Erickson)
 
+* SOLR-9716: RecoveryStrategy sends prep recovery command without setting read time out which can cause
+  replica recovery to hang indefinitely on network partitions. (Cao Manh Dat, shalin)
+
 Other Changes
 ----------------------
 

http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/1f1990d8/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java
----------------------------------------------------------------------
diff --git a/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java b/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java
index 90e515a..02b7cbd 100644
--- a/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java
+++ b/solr/core/src/java/org/apache/solr/cloud/RecoveryStrategy.java
@@ -19,6 +19,7 @@ package org.apache.solr.cloud;
 import java.io.Closeable;
 import java.io.IOException;
 import java.lang.invoke.MethodHandles;
+import java.net.SocketTimeoutException;
 import java.util.ArrayList;
 import java.util.Arrays;
 import java.util.Collections;
@@ -572,24 +573,44 @@ public class RecoveryStrategy extends Thread implements Closeable {
   private void sendPrepRecoveryCmd(String leaderBaseUrl, String leaderCoreName, Slice slice)
       throws SolrServerException, IOException, InterruptedException, ExecutionException {
 
-    try (HttpSolrClient client = new HttpSolrClient.Builder(leaderBaseUrl).build()) {
-      client.setConnectionTimeout(30000);
-      WaitForState prepCmd = new WaitForState();
-      prepCmd.setCoreName(leaderCoreName);
-      prepCmd.setNodeName(zkController.getNodeName());
-      prepCmd.setCoreNodeName(coreZkNodeName);
-      prepCmd.setState(Replica.State.RECOVERING);
-      prepCmd.setCheckLive(true);
-      prepCmd.setOnlyIfLeader(true);
-      final Slice.State state = slice.getState();
-      if (state != Slice.State.CONSTRUCTION && state != Slice.State.RECOVERY && state != Slice.State.RECOVERY_FAILED) {
-        prepCmd.setOnlyIfLeaderActive(true);
+    WaitForState prepCmd = new WaitForState();
+    prepCmd.setCoreName(leaderCoreName);
+    prepCmd.setNodeName(zkController.getNodeName());
+    prepCmd.setCoreNodeName(coreZkNodeName);
+    prepCmd.setState(Replica.State.RECOVERING);
+    prepCmd.setCheckLive(true);
+    prepCmd.setOnlyIfLeader(true);
+    final Slice.State state = slice.getState();
+    if (state != Slice.State.CONSTRUCTION && state != Slice.State.RECOVERY && state != Slice.State.RECOVERY_FAILED) {
+      prepCmd.setOnlyIfLeaderActive(true);
+    }
+
+    final int maxTries = 30;
+    for (int numTries = 0; numTries < maxTries; numTries++) {
+      try {
+        sendPrepRecoveryCmd(leaderBaseUrl, prepCmd);
+        break;
+      } catch (ExecutionException e) {
+        SolrServerException solrException = (SolrServerException) e.getCause();
+        if (solrException.getRootCause() instanceof SocketTimeoutException && numTries < maxTries) {
+          LOG.warn("Socket timeout when send prep recovery cmd, retrying.. ");
+          continue;
+        }
+        throw  e;
       }
+    }
+  }
+
+  private void sendPrepRecoveryCmd(String leaderBaseUrl, WaitForState prepCmd)
+      throws SolrServerException, IOException, InterruptedException, ExecutionException {
+    try (HttpSolrClient client = new HttpSolrClient.Builder(leaderBaseUrl).build()) {
+      client.setConnectionTimeout(10000);
+      client.setSoTimeout(10000);
       HttpUriRequestResponse mrr = client.httpUriRequest(prepCmd);
       prevSendPreRecoveryHttpUriRequest = mrr.httpUriRequest;
-      
+
       LOG.info("Sending prep recovery command to [{}]; [{}]", leaderBaseUrl, prepCmd.toString());
-      
+
       mrr.future.get();
     }
   }

http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/1f1990d8/solr/core/src/java/org/apache/solr/handler/admin/PrepRecoveryOp.java
----------------------------------------------------------------------
diff --git a/solr/core/src/java/org/apache/solr/handler/admin/PrepRecoveryOp.java b/solr/core/src/java/org/apache/solr/handler/admin/PrepRecoveryOp.java
index 45cfeb1..d370bef 100644
--- a/solr/core/src/java/org/apache/solr/handler/admin/PrepRecoveryOp.java
+++ b/solr/core/src/java/org/apache/solr/handler/admin/PrepRecoveryOp.java
@@ -37,6 +37,7 @@ import org.apache.solr.request.LocalSolrQueryRequest;
 import org.apache.solr.search.SolrIndexSearcher;
 import org.apache.solr.update.CommitUpdateCommand;
 import org.apache.solr.util.RefCounted;
+import org.apache.solr.util.TestInjection;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
@@ -46,6 +47,8 @@ class PrepRecoveryOp implements CoreAdminHandler.CoreAdminOp {
 
   @Override
   public void execute(CallInfo it) throws Exception {
+    assert TestInjection.injectPrepRecoveryOpPauseForever();
+
     final SolrParams params = it.req.getParams();
 
     String cname = params.get(CoreAdminParams.CORE);

http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/1f1990d8/solr/core/src/java/org/apache/solr/util/TestInjection.java
----------------------------------------------------------------------
diff --git a/solr/core/src/java/org/apache/solr/util/TestInjection.java b/solr/core/src/java/org/apache/solr/util/TestInjection.java
index efd80bf..0d0c023 100644
--- a/solr/core/src/java/org/apache/solr/util/TestInjection.java
+++ b/solr/core/src/java/org/apache/solr/util/TestInjection.java
@@ -24,6 +24,7 @@ import java.util.Random;
 import java.util.Set;
 import java.util.Timer;
 import java.util.TimerTask;
+import java.util.concurrent.atomic.AtomicInteger;
 import java.util.regex.Matcher;
 import java.util.regex.Pattern;
 
@@ -110,6 +111,8 @@ public class TestInjection {
   
   public static String updateRandomPause = null;
 
+  public static String prepRecoveryOpPauseForever = null;
+
   public static String randomDelayInCoreCreation = null;
   
   public static int randomDelayMaxInCoreCreationInSec = 10;
@@ -118,6 +121,8 @@ public class TestInjection {
   
   private static Set<Timer> timers = Collections.synchronizedSet(new HashSet<Timer>());
 
+  private static AtomicInteger countPrepRecoveryOpPauseForever = new AtomicInteger(0);
+
   public static void reset() {
     nonGracefullClose = null;
     failReplicaRequests = null;
@@ -127,6 +132,8 @@ public class TestInjection {
     updateRandomPause = null;
     randomDelayInCoreCreation = null;
     splitFailureBeforeReplicaCreation = null;
+    prepRecoveryOpPauseForever = null;
+    countPrepRecoveryOpPauseForever = new AtomicInteger(0);
 
     for (Timer timer : timers) {
       timer.cancel();
@@ -289,6 +296,31 @@ public class TestInjection {
     return true;
   }
 
+  public static boolean injectPrepRecoveryOpPauseForever() {
+    if (prepRecoveryOpPauseForever != null)  {
+      Random rand = random();
+      if (null == rand) return true;
+
+      Pair<Boolean,Integer> pair = parseValue(prepRecoveryOpPauseForever);
+      boolean enabled = pair.first();
+      int chanceIn100 = pair.second();
+      // Prevent for continuous pause forever
+      if (enabled && rand.nextInt(100) >= (100 - chanceIn100) && countPrepRecoveryOpPauseForever.get() < 2) {
+        countPrepRecoveryOpPauseForever.incrementAndGet();
+        log.info("inject pause forever for prep recovery op");
+        try {
+          Thread.sleep(Integer.MAX_VALUE);
+        } catch (InterruptedException e) {
+          Thread.currentThread().interrupt();
+        }
+      } else {
+        countPrepRecoveryOpPauseForever.set(0);
+      }
+    }
+
+    return true;
+  }
+
   public static boolean injectSplitFailureBeforeReplicaCreation() {
     if (splitFailureBeforeReplicaCreation != null)  {
       Random rand = random();

http://git-wip-us.apache.org/repos/asf/lucene-solr/blob/1f1990d8/solr/core/src/test/org/apache/solr/cloud/TestCloudRecovery.java
----------------------------------------------------------------------
diff --git a/solr/core/src/test/org/apache/solr/cloud/TestCloudRecovery.java b/solr/core/src/test/org/apache/solr/cloud/TestCloudRecovery.java
index 2a7413c..e2f3bfd 100644
--- a/solr/core/src/test/org/apache/solr/cloud/TestCloudRecovery.java
+++ b/solr/core/src/test/org/apache/solr/cloud/TestCloudRecovery.java
@@ -37,6 +37,8 @@ import org.apache.solr.common.params.ModifiableSolrParams;
 import org.apache.solr.core.SolrCore;
 import org.apache.solr.update.DirectUpdateHandler2;
 import org.apache.solr.update.UpdateLog;
+import org.apache.solr.util.TestInjection;
+import org.junit.AfterClass;
 import org.junit.Before;
 import org.junit.BeforeClass;
 import org.junit.Test;
@@ -47,6 +49,7 @@ public class TestCloudRecovery extends SolrCloudTestCase {
 
   @BeforeClass
   public static void setupCluster() throws Exception {
+    TestInjection.prepRecoveryOpPauseForever = "true:30";
     System.setProperty("solr.directoryFactory", "solr.StandardDirectoryFactory");
     System.setProperty("solr.ulog.numRecordsToKeep", "1000");
 
@@ -62,6 +65,11 @@ public class TestCloudRecovery extends SolrCloudTestCase {
         false, true, 30);
   }
 
+  @AfterClass
+  public static void afterClass() {
+    TestInjection.reset();
+  }
+
   @Before
   public void resetCollection() throws IOException, SolrServerException {
     cluster.getSolrClient().deleteByQuery(COLLECTION, "*:*");