[
https://issues.apache.org/jira/browse/HIVE-26509?focusedWorklogId=812337&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-812337
]
ASF GitHub Bot logged work on HIVE-26509:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 27/Sep/22 03:10
Start Date: 27/Sep/22 03:10
Worklog Time Spent: 10m
Work Description: dengzhhu653 commented on code in PR #3567:
URL: https://github.com/apache/hive/pull/3567#discussion_r980689007
##########
standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/leader/LeaseLeaderElection.java:
##########
@@ -0,0 +1,473 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+package org.apache.hadoop.hive.metastore.leader;
+
+import org.apache.hadoop.conf.Configuration;
+import org.apache.hadoop.hive.common.TableName;
+import org.apache.hadoop.hive.metastore.LockComponentBuilder;
+import org.apache.hadoop.hive.metastore.api.CheckLockRequest;
+import org.apache.hadoop.hive.metastore.api.DataOperationType;
+import org.apache.hadoop.hive.metastore.api.HeartbeatRequest;
+import org.apache.hadoop.hive.metastore.api.LockComponent;
+import org.apache.hadoop.hive.metastore.api.LockRequest;
+import org.apache.hadoop.hive.metastore.api.LockResponse;
+import org.apache.hadoop.hive.metastore.api.LockState;
+import org.apache.hadoop.hive.metastore.api.LockType;
+import org.apache.hadoop.hive.metastore.api.MetaException;
+import org.apache.hadoop.hive.metastore.api.NoSuchLockException;
+import org.apache.hadoop.hive.metastore.api.NoSuchTxnException;
+import org.apache.hadoop.hive.metastore.api.TxnAbortedException;
+import org.apache.hadoop.hive.metastore.api.TxnOpenException;
+import org.apache.hadoop.hive.metastore.api.UnlockRequest;
+import org.apache.hadoop.hive.metastore.conf.MetastoreConf;
+import org.apache.hadoop.hive.metastore.txn.TxnStore;
+import org.apache.hadoop.hive.metastore.txn.TxnUtils;
+import org.apache.hadoop.hive.metastore.utils.SecurityUtils;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.net.InetAddress;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicLong;
+
+import static java.util.Objects.requireNonNull;
+
+/**
+ * Hive Lock based leader election.
+ * If wins, the current instance becomes the leader,
+ * and a heartbeat daemon will be started to renew the lock before timeout.
+ * If loses, a non-leader watcher will also be started to check the
+ * lock periodically to see if he can grab the lock in order to be the leader.
+ * The change of Leadership can be received by registering the
+ * listeners through {@link LeaderElection#addStateListener}.
+ */
+public class LeaseLeaderElection implements LeaderElection<TableName> {
+
+ private static final Logger LOG =
LoggerFactory.getLogger(LeaseLeaderElection.class);
+
+ private static final AtomicLong ID = new AtomicLong();
+
+ // Result of election
+ private volatile boolean isLeader;
+
+ private TxnStore store;
+
+ // Initial sleep time for locking the table at retrying.
+ private long nextSleep = 50;
+
+ // A daemon used for renewing the lock before timeout,
+ // this happens when the current instance wins the election.
+ private LeaseWatcher heartbeater;
+
+ // For non-leader instances to check the lock periodically to
+ // see if there is a chance to take over the leadership.
+ // At any time, only one of heartbeater and nonLeaderWatcher is alive.
+ private LeaseWatcher nonLeaderWatcher;
+
+ // Current lock id
+ private volatile long lockId = -1;
+
+ // Leadership change listeners
+ private List<LeadershipStateListener> listeners = new ArrayList<>();
+
+ // Property for testing only
+ public static final String METASTORE_RENEW_LEASE =
"metastore.renew.leader.lease";
Review Comment:
The property is just for testing now, for making the lock timeout by no
heartbeat and trying to acquire the lock again, so others in the tests have a
chance to grab the lock to become a leader, in this case, the original leader
should suspend the tasks when transmitting to non-leader.
Issue Time Tracking
-------------------
Worklog Id: (was: 812337)
Time Spent: 4h (was: 3h 50m)
> Introduce dynamic leader election in HMS
> ----------------------------------------
>
> Key: HIVE-26509
> URL: https://issues.apache.org/jira/browse/HIVE-26509
> Project: Hive
> Issue Type: New Feature
> Components: Standalone Metastore
> Reporter: Zhihua Deng
> Priority: Major
> Labels: pull-request-available
> Time Spent: 4h
> Remaining Estimate: 0h
>
> From HIVE-21841 we have a leader HMS selected by configuring
> metastore.housekeeping.leader.hostname on startup. This approach saves us
> from running duplicated HMS's housekeeping tasks cluster-wide.
> In this jira, we introduce another dynamic leader election: adopt hive lock
> to implement the leader election. Once a HMS owns the lock, then it becomes
> the leader, carries out the housekeeping tasks, and sends heartbeats to renew
> the lock before timeout. If the leader fails to reclaim the lock, then stops
> the already started tasks if it has, the electing event is audited. We can
> achieve a more dynamic leader when the original goes down or in the public
> cloud without well configured property, and reduce the leader’s burdens by
> running these tasks among different leaders.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
