Merge pull request #6 from linkedin/master

Release for helix 1.3.2-dev-202406121430

helix-common/src/main/java/org/apache/helix/constants/InstanceConstants.java

@@ -1,17 +1,103 @@
 package org.apache.helix.constants;
 
+import java.util.Set;
+
+import com.google.common.collect.ImmutableSet;
+
 public class InstanceConstants {
   public static final String INSTANCE_NOT_DISABLED = "INSTANCE_NOT_DISABLED";
 
+  /**
+   * The set contains the InstanceOperations that are allowed to be assigned replicas by the rebalancer.
+   */
+  public static final Set<InstanceOperation> ASSIGNABLE_INSTANCE_OPERATIONS =
+      ImmutableSet.of(InstanceOperation.ENABLE, InstanceOperation.DISABLE);
+
+
+  /**
+   * The set contains the InstanceOperations that are overridden when the deprecated HELIX_ENABLED
+   * field is set to false. This will maintain backwards compatibility with the deprecated field.
+   * TODO: Remove this when the deprecated HELIX_ENABLED is removed.
+   */
+  public static final Set<InstanceOperation> INSTANCE_DISABLED_OVERRIDABLE_OPERATIONS =
+      ImmutableSet.of(InstanceOperation.ENABLE, InstanceOperation.EVACUATE);
+
+
+  /**
+   * The set of InstanceOperations that are not allowed to be populated in the RoutingTableProvider.
+   */
+  public static final Set<InstanceOperation> UNROUTABLE_INSTANCE_OPERATIONS =
+      ImmutableSet.of(InstanceOperation.SWAP_IN, InstanceOperation.UNKNOWN);
+
+  @Deprecated
   public enum InstanceDisabledType {
     CLOUD_EVENT,
     USER_OPERATION,
     DEFAULT_INSTANCE_DISABLE_TYPE
   }
 
+  public enum InstanceOperationSource {
+    ADMIN(0), USER(1), AUTOMATION(2), DEFAULT(3);
+
+    private final int _priority;
+
+    InstanceOperationSource(int priority) {
+      _priority = priority;
+    }
+
+    public int getPriority() {
+      return _priority;
+    }
+
+    /**
+     * Convert from InstanceDisabledType to InstanceOperationTrigger
+     *
+     * @param disabledType InstanceDisabledType
+     * @return InstanceOperationTrigger
+     */
+    public static InstanceOperationSource instanceDisabledTypeToInstanceOperationSource(
+        InstanceDisabledType disabledType) {
+      switch (disabledType) {
+        case CLOUD_EVENT:
+          return InstanceOperationSource.AUTOMATION;
+        case USER_OPERATION:
+          return InstanceOperationSource.USER;
+        default:
+          return InstanceOperationSource.DEFAULT;
+      }
+    }
+  }
+
   public enum InstanceOperation {
-    EVACUATE, // Node will be removed after a period of time
-    SWAP_IN,  // New node joining for swap operation
-    SWAP_OUT // Existing Node to be removed for swap operation
+    /**
+     * Behavior: Replicas will be assigned to the node and will receive upward state transitions if
+     * for new assignments and downward state transitions if replicas are being moved elsewhere.
+     * Final State: The node will have replicas assigned to it and will be considered for future assignment.
+     */
+    ENABLE,
+    /**
+     * Behavior: All replicas on the node will be set to OFFLINE.
+     * Final State: The node will have all replicas in the OFFLINE state and can't take new assignment.
+     */
+    DISABLE,
+    /**
+     * Behavior: All replicas will be moved off the node, after a replacement has been bootstrapped
+     * in another node in the cluster.
+     * Final State: The node will not contain any replicas and will not be considered for *NEW* assignment.
+     */
+    EVACUATE,
+    /**
+     * Behavior: Node will have all replicas on its corresponding(same logicalId) swap-out node bootstrapped
+     * (ERROR and OFFLINE replicas on swap-out node will not be bootstrapped) to the same states if the StateModelDef allows.
+     * This node will be excluded from the RoutingTableProvider.
+     * Final State: This node will be a mirror the swap-out node, will not be considered for assignment, and will not be populated
+     * in the RoutingTableProvider.
+     */
+    SWAP_IN,
+    /**
+     * Behavior: Node will have all of its replicas dropped immediately and will be removed from the RoutingTableProvider.
+     * Final State: Node will not hold replicas, be considered for assignment, or be populated in the RoutingTableProvider.
+     */
+    UNKNOWN
   }
 }

helix-core/src/main/java/org/apache/helix/HelixAdmin.java

@@ -281,9 +281,11 @@ void addResource(String clusterName, String resourceName, int numPartitions, Str
    * @param instanceName
    * @param enabled
    */
+  @Deprecated
   void enableInstance(String clusterName, String instanceName, boolean enabled);
 
   /**
+   * @deprecated use {@link #setInstanceOperation(String, String, InstanceConstants.InstanceOperation)}
    * @param clusterName
    * @param instanceName
    * @param enabled
@@ -292,27 +294,54 @@ void addResource(String clusterName, String resourceName, int numPartitions, Str
    * @param reason set additional string description on why the instance is disabled when
    *          <code>enabled</code> is false. Existing disabled reason will be over write if instance is in disabled state.
    */
+  @Deprecated
   void enableInstance(String clusterName, String instanceName, boolean enabled,
       InstanceConstants.InstanceDisabledType disabledType, String reason);
 
   /**
    * Batch enable/disable instances in a cluster
    * By default, all the instances are enabled
+   * @deprecated use {@link #setInstanceOperation(String, String, InstanceConstants.InstanceOperation)}
    * @param clusterName
    * @param instances
    * @param enabled
    */
+  @Deprecated
   void enableInstance(String clusterName, List<String> instances, boolean enabled);
 
   /**
-   * Set the instanceOperation field.
+   * Set the instanceOperation of and instance with {@link InstanceConstants.InstanceOperation}.
    *
    * @param clusterName       The cluster name
    * @param instanceName      The instance name
-   * @param instanceOperation The instance operation
+   * @param instanceOperation The instance operation type
    */
   void setInstanceOperation(String clusterName, String instanceName,
-      @Nullable InstanceConstants.InstanceOperation instanceOperation);
+      InstanceConstants.InstanceOperation instanceOperation);
+
+  /**
+   * Set the instanceOperation of and instance with {@link InstanceConstants.InstanceOperation}.
+   *
+   * @param clusterName       The cluster name
+   * @param instanceName      The instance name
+   * @param instanceOperation The instance operation type
+   * @param reason            The reason for the operation
+   */
+  void setInstanceOperation(String clusterName, String instanceName,
+      InstanceConstants.InstanceOperation instanceOperation, String reason);
+
+  /**
+   * Set the instanceOperation of and instance with {@link InstanceConstants.InstanceOperation}.
+   *
+   * @param clusterName       The cluster name
+   * @param instanceName      The instance name
+   * @param instanceOperation The instance operation type
+   * @param reason            The reason for the operation
+   * @param overrideAll       Whether to override all existing instance operations from all other
+   *                          instance operations
+   */
+  void setInstanceOperation(String clusterName, String instanceName,
+      InstanceConstants.InstanceOperation instanceOperation, String reason, boolean overrideAll);
 
   /**
    * Disable or enable a resource
@@ -415,6 +444,18 @@ void manuallyEnableMaintenanceMode(String clusterName, boolean enabled, String r
    */
   ClusterManagementMode getClusterManagementMode(String clusterName);
 
+  /**
+   * Set a list of partitions for an instance to ERROR state from any state.
+   * The partitions could be in any state and setPartitionsToError will bring them to ERROR
+   * state. ANY to ERROR state transition is required for this.
+   * @param clusterName
+   * @param instanceName
+   * @param resourceName
+   * @param partitionNames
+   */
+  void setPartitionsToError(String clusterName, String instanceName, String resourceName,
+      List<String> partitionNames);
+
   /**
    * Reset a list of partitions in error state for an instance
    * The partitions are assume to be in error state and reset will bring them from error

helix-core/src/main/java/org/apache/helix/cloud/event/helix/DefaultCloudEventCallbackImpl.java

@@ -23,6 +23,8 @@
 import org.apache.helix.HelixManager;
 import org.apache.helix.constants.InstanceConstants;
 import org.apache.helix.model.ClusterConfig;
+import org.apache.helix.model.InstanceConfig;
+import org.apache.helix.util.InstanceUtil;
 import org.apache.helix.util.InstanceValidationUtil;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -49,9 +51,14 @@ public void disableInstance(HelixManager manager, Object eventInfo) {
     LOG.info("DefaultCloudEventCallbackImpl disable Instance {}", manager.getInstanceName());
     if (InstanceValidationUtil
         .isEnabled(manager.getHelixDataAccessor(), manager.getInstanceName())) {
-      manager.getClusterManagmentTool()
-          .enableInstance(manager.getClusterName(), manager.getInstanceName(), false,
-              InstanceConstants.InstanceDisabledType.CLOUD_EVENT, message);
+      InstanceUtil.setInstanceOperation(manager.getConfigAccessor(),
+          manager.getHelixDataAccessor().getBaseDataAccessor(), manager.getClusterName(),
+          manager.getInstanceName(),
+              new InstanceConfig.InstanceOperation.Builder().setOperation(
+                      InstanceConstants.InstanceOperation.DISABLE)
+                  .setSource(InstanceConstants.InstanceOperationSource.AUTOMATION)
+                  .setReason(message)
+                  .build());
     }
     HelixEventHandlingUtil.updateCloudEventOperationInClusterConfig(manager.getClusterName(),
         manager.getInstanceName(), manager.getHelixDataAccessor().getBaseDataAccessor(), false,
@@ -72,10 +79,13 @@ public void enableInstance(HelixManager manager, Object eventInfo) {
     HelixEventHandlingUtil
         .updateCloudEventOperationInClusterConfig(manager.getClusterName(), instanceName,
             manager.getHelixDataAccessor().getBaseDataAccessor(), true, message);
-    if (HelixEventHandlingUtil.isInstanceDisabledForCloudEvent(instanceName, accessor)) {
-      manager.getClusterManagmentTool().enableInstance(manager.getClusterName(), instanceName, true,
-          InstanceConstants.InstanceDisabledType.CLOUD_EVENT, message);
-    }
+    InstanceUtil.setInstanceOperation(manager.getConfigAccessor(),
+        manager.getHelixDataAccessor().getBaseDataAccessor(), manager.getClusterName(),
+        manager.getInstanceName(),
+            new InstanceConfig.InstanceOperation.Builder().setOperation(
+                    InstanceConstants.InstanceOperation.ENABLE)
+                .setSource(InstanceConstants.InstanceOperationSource.AUTOMATION).setReason(message)
+                .build());
   }
 
   /**

helix-core/src/main/java/org/apache/helix/cloud/event/helix/HelixEventHandlingUtil.java

@@ -48,7 +48,10 @@ class HelixEventHandlingUtil {
    * @param dataAccessor
    * @return return true only when instance is Helix disabled and the disabled reason in
    * instanceConfig is cloudEvent
+   * @deprecated No need to check this if using InstanceOperation and specifying the trigger as CLOUD
+   *            when enabling.
    */
+  @Deprecated
   static boolean isInstanceDisabledForCloudEvent(String instanceName,
       HelixDataAccessor dataAccessor) {
     InstanceConfig instanceConfig =

helix-core/src/main/java/org/apache/helix/controller/changedetector/trimmer/InstanceConfigTrimmer.java

@@ -19,6 +19,7 @@
  * under the License.
  */
 
+import java.util.HashMap;
 import java.util.Map;
 import java.util.Set;
 
@@ -62,6 +63,21 @@ protected Map<FieldType, Set<String>> getNonTrimmableFields(InstanceConfig insta
     return STATIC_TOPOLOGY_RELATED_FIELD_MAP;
   }
 
+  /**
+   * We should trim HELIX_INSTANCE_OPERATIONS field, it is used to filter instances in the
+   * BaseControllerDataProvider. That filtering will be used to determine if ResourceChangeSnapshot
+   * has changed as opposed to checking the actual value of the field.
+   *
+   * @param property the instance config
+   * @return a map contains all non-trimmable field keys that need to be kept.
+   */
+  protected Map<FieldType, Set<String>> getNonTrimmableKeys(InstanceConfig property) {
+    Map<FieldType, Set<String>> nonTrimmableKeys = super.getNonTrimmableKeys(property);
+    nonTrimmableKeys.get(FieldType.LIST_FIELD)
+        .remove(InstanceConfigProperty.HELIX_INSTANCE_OPERATIONS.name());
+    return nonTrimmableKeys;
+  }
+
   @Override
   public InstanceConfig trimProperty(InstanceConfig property) {
     return new InstanceConfig(doTrim(property));