[feature](merge-cloud) Add tablet rebalance run in cloud

apache · Mar 30, 2024 · d7a152f · d7a152f
1 parent 2b94cd1
commit d7a152f
Show file tree

Hide file tree

Showing 8 changed files with 1,233 additions and 17 deletions.
diff --git a/fe/fe-common/src/main/java/org/apache/doris/common/Config.java b/fe/fe-common/src/main/java/org/apache/doris/common/Config.java
@@ -2635,6 +2635,35 @@ public static boolean isNotCloudMode() {
     @ConfField
     public static int drop_user_notify_ms_max_times = 86400;
 
+    @ConfField(mutable = true)
+    public static long cloud_tablet_rebalancer_interval_second = 20;
+
+    @ConfField(mutable = true)
+    public static boolean enable_cloud_partition_balance = true;
+
+    @ConfField(mutable = true)
+    public static boolean enable_cloud_table_balance = true;
+
+    @ConfField(mutable = true)
+    public static boolean enable_cloud_global_balance = true;
+
+    @ConfField(mutable = true)
+    public static int cloud_pre_heating_time_limit_sec = 300;
+
+    @ConfField(mutable = true)
+    public static double cloud_rebalance_percent_threshold = 0.05;
+
+    @ConfField(mutable = true)
+    public static long cloud_rebalance_number_threshold = 2;
+
+    @ConfField(mutable = true)
+    public static double cloud_balance_tablet_percent_per_run = 0.05;
+
+    @ConfField(mutable = true)
+    public static int cloud_min_balance_tablet_num_per_run = 2;
+
+    @ConfField(mutable = true)
+    public static boolean cloud_preheating_enabled = true;
     //==========================================================================
     //                      end of cloud config
     //==========================================================================

diff --git a/fe/fe-core/src/main/java/org/apache/doris/catalog/Env.java b/fe/fe-core/src/main/java/org/apache/doris/catalog/Env.java
@@ -1646,24 +1646,26 @@ protected void startMasterOnlyDaemonThreads() {
         loadJobScheduler.start();
         loadEtlChecker.start();
         loadLoadingChecker.start();
-        // Tablet checker and scheduler
-        tabletChecker.start();
-        tabletScheduler.start();
-        // Colocate tables checker and balancer
-        ColocateTableCheckerAndBalancer.getInstance().start();
-        // Publish Version Daemon
-        publishVersionDaemon.start();
-        // Start txn cleaner
-        txnCleaner.start();
+        if (Config.isNotCloudMode()) {
+            // Tablet checker and scheduler
+            tabletChecker.start();
+            tabletScheduler.start();
+            // Colocate tables checker and balancer
+            ColocateTableCheckerAndBalancer.getInstance().start();
+            // Publish Version Daemon
+            publishVersionDaemon.start();
+            // Start txn cleaner
+            txnCleaner.start();
+            // Consistency checker
+            getConsistencyChecker().start();
+            // Backup handler
+            getBackupHandler().start();
+        }
         jobManager.start();
         // transient task manager
         transientTaskManager.start();
         // Alter
         getAlterInstance().start();
-        // Consistency checker
-        getConsistencyChecker().start();
-        // Backup handler
-        getBackupHandler().start();
         // catalog recycle bin
         getRecycleBin().start();
         // time printer

diff --git a/fe/fe-core/src/main/java/org/apache/doris/cloud/catalog/CloudEnv.java b/fe/fe-core/src/main/java/org/apache/doris/cloud/catalog/CloudEnv.java
@@ -19,12 +19,15 @@
 
 import org.apache.doris.analysis.ResourceTypeEnum;
 import org.apache.doris.catalog.Env;
+import org.apache.doris.cloud.datasource.CloudInternalCatalog;
+import org.apache.doris.cloud.persist.UpdateCloudReplicaInfo;
 import org.apache.doris.cloud.proto.Cloud;
 import org.apache.doris.cloud.proto.Cloud.NodeInfoPB;
 import org.apache.doris.cloud.system.CloudSystemInfoService;
 import org.apache.doris.common.Config;
 import org.apache.doris.common.DdlException;
 import org.apache.doris.common.ErrorCode;
+import org.apache.doris.common.MetaNotFoundException;
 import org.apache.doris.common.UserException;
 import org.apache.doris.common.io.CountingDataOutputStream;
 import org.apache.doris.common.util.HttpURLUtil;
@@ -57,17 +60,25 @@ public class CloudEnv extends Env {
     private CloudInstanceStatusChecker cloudInstanceStatusChecker;
     private CloudClusterChecker cloudClusterCheck;
 
+    private CloudTabletRebalancer cloudTabletRebalancer;
+
     public CloudEnv(boolean isCheckpointCatalog) {
         super(isCheckpointCatalog);
         this.cloudClusterCheck = new CloudClusterChecker((CloudSystemInfoService) systemInfo);
         this.cloudInstanceStatusChecker = new CloudInstanceStatusChecker((CloudSystemInfoService) systemInfo);
+        this.cloudTabletRebalancer = new CloudTabletRebalancer((CloudSystemInfoService) systemInfo);
+    }
+
+    public CloudTabletRebalancer getCloudTabletRebalancer() {
+        return this.cloudTabletRebalancer;
     }
 
     @Override
     protected void startMasterOnlyDaemonThreads() {
         LOG.info("start cloud Master only daemon threads");
         super.startMasterOnlyDaemonThreads();
         cloudClusterCheck.start();
+        cloudTabletRebalancer.start();
     }
 
     @Override
@@ -416,4 +427,8 @@ public String analyzeCloudCluster(String name, ConnectContext ctx) throws DdlExc
         changeCloudCluster(res[1], ctx);
         return res[0];
     }
+
+    public void replayUpdateCloudReplica(UpdateCloudReplicaInfo info) throws MetaNotFoundException {
+        ((CloudInternalCatalog) getInternalCatalog()).replayUpdateCloudReplica(info);
+    }
 }