From 57c35ad340e376c75b17b4ca7ffbd91376f10963 Mon Sep 17 00:00:00 2001
From: Xianming Lei <31424839+leixm@users.noreply.github.com>
Date: Fri, 28 Jul 2023 14:05:04 +0800
Subject: [PATCH] [#1048] fix(coord): ExcludeNodes does not take effect when
 the coordinator restarts. (#1049)

### What changes were proposed in this pull request?

Make updateExcludeNodes no delay first time.

### Why are the changes needed?

Fix: #1048.

### Does this PR introduce _any_ user-facing change?

No.

### How was this patch tested?

UT.

Co-authored-by: leixianming <leixianming@didiglobal.com>
---
 .../coordinator/SimpleClusterManager.java     |  2 +-
 .../coordinator/SimpleClusterManagerTest.java | 27 +++++++++++++++++++
 2 files changed, 28 insertions(+), 1 deletion(-)

diff --git a/coordinator/src/main/java/org/apache/uniffle/coordinator/SimpleClusterManager.java b/coordinator/src/main/java/org/apache/uniffle/coordinator/SimpleClusterManager.java
index 8c5929170c..4ea6ee0d09 100644
--- a/coordinator/src/main/java/org/apache/uniffle/coordinator/SimpleClusterManager.java
+++ b/coordinator/src/main/java/org/apache/uniffle/coordinator/SimpleClusterManager.java
@@ -114,7 +114,7 @@ public SimpleClusterManager(CoordinatorConf conf, Configuration hadoopConf) thro
           ThreadUtils.getDaemonSingleThreadScheduledExecutor("UpdateExcludeNodes");
       checkNodesExecutorService.scheduleAtFixedRate(
           () -> updateExcludeNodes(excludeNodesPath),
-          updateNodesInterval,
+          0,
           updateNodesInterval,
           TimeUnit.MILLISECONDS);
     }
diff --git a/coordinator/src/test/java/org/apache/uniffle/coordinator/SimpleClusterManagerTest.java b/coordinator/src/test/java/org/apache/uniffle/coordinator/SimpleClusterManagerTest.java
index b3502dcf6f..1391308f8b 100644
--- a/coordinator/src/test/java/org/apache/uniffle/coordinator/SimpleClusterManagerTest.java
+++ b/coordinator/src/test/java/org/apache/uniffle/coordinator/SimpleClusterManagerTest.java
@@ -28,6 +28,7 @@
 import java.util.stream.Collectors;
 
 import com.google.common.collect.Sets;
+import com.google.common.util.concurrent.Uninterruptibles;
 import org.apache.commons.collections.CollectionUtils;
 import org.apache.hadoop.conf.Configuration;
 import org.junit.jupiter.api.AfterEach;
@@ -487,6 +488,32 @@ public void updateExcludeNodesTest() throws Exception {
     }
   }
 
+  @Test
+  public void excludeNodesNoDelayTest() throws Exception {
+    String excludeNodesFolder =
+        (new File(ClassLoader.getSystemResource("empty").getFile())).getParent();
+    String excludeNodesPath = excludeNodesFolder + "/excludeNodes";
+    CoordinatorConf ssc = new CoordinatorConf();
+    ssc.setString(
+        CoordinatorConf.COORDINATOR_EXCLUDE_NODES_FILE_PATH,
+        URI.create(excludeNodesPath).toString());
+    ssc.setLong(CoordinatorConf.COORDINATOR_EXCLUDE_NODES_CHECK_INTERVAL, 5000);
+
+    final Set<String> nodes = Sets.newHashSet("node1-1999", "node2-1999");
+    writeExcludeHosts(excludeNodesPath, nodes);
+
+    try (SimpleClusterManager scm = new SimpleClusterManager(ssc, new Configuration())) {
+      // waiting for excludeNode file parse.
+      Uninterruptibles.sleepUninterruptibly(20, TimeUnit.MILLISECONDS);
+      scm.add(new ServerNode("node1-1999", "ip", 0, 100L, 50L, 20, 10, testTags));
+      scm.add(new ServerNode("node2-1999", "ip", 0, 100L, 50L, 20, 10, testTags));
+      scm.add(new ServerNode("node3-1999", "ip", 0, 100L, 50L, 20, 10, testTags));
+      scm.add(new ServerNode("node4-1999", "ip", 0, 100L, 50L, 20, 10, testTags));
+      assertEquals(4, scm.getNodesNum());
+      assertEquals(2, scm.getExcludeNodes().size());
+    }
+  }
+
   private void writeExcludeHosts(String path, Set<String> values) throws Exception {
     try (PrintWriter pw = new PrintWriter(new FileWriter(path))) {
       // have empty line as value