SOLR-9280 - make nodeName a configurable parameter in solr.xml

solr/core/src/java/org/apache/solr/cloud/ZkController.java

@@ -181,7 +181,8 @@ public boolean equals(Object obj) {
 
   private final int localHostPort;      // example: 54065
   private final String hostName;           // example: 127.0.0.1
-  private final String nodeName;           // example: 127.0.0.1:54065_solr
+  private final String genericNodeName;           // example: 127.0.0.1:54065_solr
+  private final String nodeName;          //set in solr.xml or genericNodeName as default
   private String baseURL;            // example: http://127.0.0.1:54065/solr
 
   private final CloudConfig cloudConfig;
@@ -271,8 +272,11 @@ public ZkController(final CoreContainer cc, String zkServerAddress, int zkClient
     this.zkServerAddress = zkServerAddress;
     this.localHostPort = cloudConfig.getSolrHostPort();
     this.hostName = normalizeHostName(cloudConfig.getHost());
-    this.nodeName = generateNodeName(this.hostName, Integer.toString(this.localHostPort), localHostContext);
-    MDCLoggingContext.setNode(nodeName);
+    this.genericNodeName = generateNodeName(this.hostName, Integer.toString(this.localHostPort), localHostContext);
+    String nodeName = cloudConfig.getNodeName();
+    this.nodeName = nodeName == null || nodeName.isEmpty() ? this.genericNodeName : nodeName;
+    String loggingContext = this.nodeName == this.genericNodeName ? this.genericNodeName : String.format("%s[%s]", this.nodeName, this.genericNodeName);
+    MDCLoggingContext.setNode(loggingContext);
     this.leaderVoteWait = cloudConfig.getLeaderVoteWait();
     this.leaderConflictResolveWait = cloudConfig.getLeaderConflictResolveWait();
 
@@ -671,7 +675,7 @@ private void init(CurrentCoreDescriptorProvider registerOnReconnect) {
     try {
       createClusterZkNodes(zkClient);
       zkStateReader.createClusterStateWatchersAndUpdate();
-      this.baseURL = zkStateReader.getBaseUrlForNodeName(this.nodeName);
+      this.baseURL = zkStateReader.getBaseUrlFromGenericNodeName(this.genericNodeName);
 
       checkForExistingEphemeralNode();
 
@@ -738,9 +742,10 @@ private void checkForExistingEphemeralNode() throws KeeperException, Interrupted
     }
 
     boolean deleted = deletedLatch.await(zkClient.getSolrZooKeeper().getSessionTimeout() * 2, TimeUnit.MILLISECONDS);
+
     if (!deleted) {
-      throw new SolrException(ErrorCode.SERVER_ERROR, "A previous ephemeral live node still exists. " +
-          "Solr cannot continue. Please ensure that no other Solr process using the same port is running already.");
+      throw new SolrException(ErrorCode.SERVER_ERROR, "An ephemeral live node still exists named '" + this.nodeName +
+          "'. Solr cannot continue. Please ensure that no other Solr process using the same port is running already or configured with the same nodeName in solr.xml");
     }
   }
 
@@ -819,7 +824,11 @@ private void createEphemeralLiveNode() throws KeeperException,
     String nodeName = getNodeName();
     String nodePath = ZkStateReader.LIVE_NODES_ZKNODE + "/" + nodeName;
     log.info("Register node as live in ZooKeeper:" + nodePath);
-    zkClient.makePath(nodePath, CreateMode.EPHEMERAL, true);
+    zkClient.makePath(nodePath, getGenericNodeName().getBytes(), CreateMode.EPHEMERAL, true);
+  }
+
+  private String getGenericNodeName() {
+    return genericNodeName;
   }
 
   public String getNodeName() {
@@ -2104,7 +2113,7 @@ public void updateLeaderInitiatedRecoveryState(String collection, String shardId
     stateObj.put(ZkStateReader.STATE_PROP, state.toString());
     // only update the createdBy value if it's not set
     if (stateObj.get("createdByNodeName") == null) {
-      stateObj.put("createdByNodeName", this.nodeName);
+      stateObj.put("createdByNodeName", this.genericNodeName);
     }
     if (stateObj.get("createdByCoreNodeName") == null && leaderCoreNodeName != null)  {
       stateObj.put("createdByCoreNodeName", leaderCoreNodeName);

solr/core/src/java/org/apache/solr/core/CloudConfig.java

@@ -49,11 +49,13 @@ public class CloudConfig {
   private final int createCollectionWaitTimeTillActive;
 
   private final boolean createCollectionCheckLeaderActive;
+
+  private final String nodeName;
 
   CloudConfig(String zkHost, int zkClientTimeout, int hostPort, String hostName, String hostContext, boolean useGenericCoreNames, 
               int leaderVoteWait, int leaderConflictResolveWait, int autoReplicaFailoverWaitAfterExpiration, 
               int autoReplicaFailoverWorkLoopDelay, int autoReplicaFailoverBadNodeExpiration, String zkCredentialsProviderClass, 
-              String zkACLProviderClass, int createCollectionWaitTimeTillActive, boolean createCollectionCheckLeaderActive) {
+              String zkACLProviderClass, int createCollectionWaitTimeTillActive, boolean createCollectionCheckLeaderActive, String nodeName) {
     this.zkHost = zkHost;
     this.zkClientTimeout = zkClientTimeout;
     this.hostPort = hostPort;
@@ -69,6 +71,7 @@ public class CloudConfig {
     this.zkACLProviderClass = zkACLProviderClass;
     this.createCollectionWaitTimeTillActive = createCollectionWaitTimeTillActive;
     this.createCollectionCheckLeaderActive = createCollectionCheckLeaderActive;
+    this.nodeName = nodeName;
 
     if (this.hostPort == -1)
       throw new SolrException(SolrException.ErrorCode.SERVER_ERROR, "'hostPort' must be configured to run SolrCloud");
@@ -135,6 +138,10 @@ public int getCreateCollectionWaitTimeTillActive() {
   public boolean isCreateCollectionCheckLeaderActive() {
     return createCollectionCheckLeaderActive;
   }
+
+  public String getNodeName() {
+    return nodeName;
+  }
 
   public static class CloudConfigBuilder {
 
@@ -164,6 +171,7 @@ public static class CloudConfigBuilder {
     private String zkACLProviderClass;
     private int createCollectionWaitTimeTillActive = DEFAULT_CREATE_COLLECTION_ACTIVE_WAIT;
     private boolean createCollectionCheckLeaderActive = DEFAULT_CREATE_COLLECTION_CHECK_LEADER_ACTIVE;
+    private String nodeName = null;
 
     public CloudConfigBuilder(String hostName, int hostPort) {
       this(hostName, hostPort, null);
@@ -235,11 +243,16 @@ public CloudConfigBuilder setCreateCollectionCheckLeaderActive(boolean createCol
       return this;
     }
 
+    public CloudConfigBuilder setNodeName(String nodeName) {
+      this.nodeName = nodeName;
+      return this;
+    }
+
     public CloudConfig build() {
       return new CloudConfig(zkHost, zkClientTimeout, hostPort, hostName, hostContext, useGenericCoreNames, leaderVoteWait, 
                              leaderConflictResolveWait, autoReplicaFailoverWaitAfterExpiration, autoReplicaFailoverWorkLoopDelay, 
                              autoReplicaFailoverBadNodeExpiration, zkCredentialsProviderClass, zkACLProviderClass, createCollectionWaitTimeTillActive,
-                             createCollectionCheckLeaderActive);
+                             createCollectionCheckLeaderActive, nodeName);
     }
   }
 }

solr/core/src/java/org/apache/solr/core/CoreContainer.java

@@ -1237,6 +1237,10 @@ public ZkController getZkController() {
     return zkSys.getZkController();
   }
 
+  public ZkStateReader getZkStateReader() {
+    return isZooKeeperAware() ? zkSys.getZkController().getZkStateReader() : null;
+  }
+
   public NodeConfig getConfig() {
     return cfg;
   }

solr/core/src/java/org/apache/solr/core/SolrXmlConfig.java

@@ -364,6 +364,9 @@ private static CloudConfig fillSolrCloudSection(NamedList<Object> nl) {
         case "genericCoreNodeNames":
           builder.setUseGenericCoreNames(Boolean.parseBoolean(value));
           break;
+        case "nodeName":
+          builder.setNodeName(value);
+          break;
         case "zkACLProvider":
           builder.setZkACLProviderClass(value);
           break;

solr/core/src/test-files/solr/solr.xml

@@ -37,6 +37,7 @@
     <str name="hostContext">${hostContext:solr}</str>
     <int name="zkClientTimeout">${solr.zkclienttimeout:30000}</int>
     <bool name="genericCoreNodeNames">${genericCoreNodeNames:true}</bool>
+    <str name="nodeName">${solr.nodeName:}</str>
     <int name="leaderVoteWait">${leaderVoteWait:10000}</int>
     <int name="distribUpdateConnTimeout">${distribUpdateConnTimeout:45000}</int>
     <int name="distribUpdateSoTimeout">${distribUpdateSoTimeout:340000}</int>

solr/core/src/test/org/apache/solr/cloud/HttpPartitionTest.java

@@ -635,14 +635,6 @@ private NamedList realTimeGetDocId(HttpSolrClient solr, String docId) throws Sol
     return solr.request(qr);
   }
 
-  protected int getReplicaPort(Replica replica) {
-    String replicaNode = replica.getNodeName();    
-    String tmp = replicaNode.substring(replicaNode.indexOf(':')+1);
-    if (tmp.indexOf('_') != -1)
-      tmp = tmp.substring(0,tmp.indexOf('_'));
-    return Integer.parseInt(tmp);    
-  }
-
   protected void waitToSeeReplicasActive(String testCollectionName, String shardId, Set<String> replicasToCheck, int maxWaitSecs) throws Exception {
     final RTimer timer = new RTimer();
 

solr/core/src/test/org/apache/solr/cloud/RollingRestartTest.java

@@ -66,6 +66,7 @@ public void restartWithRolesTest() throws Exception {
     numDesignateOverseers = Math.max(getShardCount(), numDesignateOverseers);
     List<String> designates = new ArrayList<>();
     List<CloudJettyRunner> designateJettys = new ArrayList<>();
+
     for (int i = 0; i < numDesignateOverseers; i++) {
       int n = random().nextInt(getShardCount());
       String nodeName = cloudJettys.get(n).nodeName;

solr/core/src/test/org/apache/solr/cloud/ZkControllerTest.java

@@ -97,41 +97,41 @@ public void testNodeNameUrlConversion() throws Exception {
 
           // getBaseUrlForNodeName
           assertEquals("http://zzz.xxx:1234/solr",
-              zkStateReader.getBaseUrlForNodeName("zzz.xxx:1234_solr"));
+              zkStateReader.getBaseUrlFromGenericNodeName("zzz.xxx:1234_solr"));
           assertEquals("http://xxx:99",
-              zkStateReader.getBaseUrlForNodeName("xxx:99_"));
+              zkStateReader.getBaseUrlFromGenericNodeName("xxx:99_"));
           assertEquals("http://foo-bar.baz.org:9999/some_dir",
-              zkStateReader.getBaseUrlForNodeName("foo-bar.baz.org:9999_some_dir"));
+              zkStateReader.getBaseUrlFromGenericNodeName("foo-bar.baz.org:9999_some_dir"));
           assertEquals("http://foo-bar.baz.org:9999/solr/sub_dir",
-              zkStateReader.getBaseUrlForNodeName("foo-bar.baz.org:9999_solr%2Fsub_dir"));
+              zkStateReader.getBaseUrlFromGenericNodeName("foo-bar.baz.org:9999_solr%2Fsub_dir"));
 
           // generateNodeName + getBaseUrlForNodeName
           assertEquals("http://foo:9876/solr",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo", "9876", "solr")));
           assertEquals("http://foo:9876/solr",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo", "9876", "/solr")));
           assertEquals("http://foo:9876/solr",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo", "9876", "/solr/")));
           assertEquals("http://foo.bar.com:9876/solr/sub_dir",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo.bar.com", "9876", "solr/sub_dir")));
           assertEquals("http://foo.bar.com:9876/solr/sub_dir",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo.bar.com", "9876", "/solr/sub_dir/")));
           assertEquals("http://foo-bar:9876",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo-bar", "9876", "")));
           assertEquals("http://foo-bar:9876",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo-bar", "9876", "/")));
           assertEquals("http://foo-bar.com:80/some_dir",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo-bar.com", "80", "some_dir")));
           assertEquals("http://foo-bar.com:80/some_dir",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo-bar.com", "80", "/some_dir")));
 
         }
@@ -146,10 +146,10 @@ public void testNodeNameUrlConversion() throws Exception {
           zkStateReader.createClusterStateWatchersAndUpdate();
 
           assertEquals("https://zzz.xxx:1234/solr",
-              zkStateReader.getBaseUrlForNodeName("zzz.xxx:1234_solr"));
+              zkStateReader.getBaseUrlFromGenericNodeName("zzz.xxx:1234_solr"));
 
           assertEquals("https://foo-bar.com:80/some_dir",
-              zkStateReader.getBaseUrlForNodeName
+              zkStateReader.getBaseUrlFromGenericNodeName
                   (ZkController.generateNodeName("foo-bar.com", "80", "/some_dir")));
 
         }