backups don't become primaries on node failures, primaries are first allocated and then start initializing from gateway

This commit is contained in:
kimchy 2010-06-29 18:20:13 +03:00
parent 514df4ee3f
commit de4b21e986
2 changed files with 107 additions and 3 deletions

View File

@ -78,14 +78,17 @@ public class ShardsRoutingStrategy {
// create a sorted list of from nodes with least number of shards to the maximum ones
applyNewNodes(routingNodes, dataNodes);
// elect primaries *before* allocating unassigned, so backups of primaries that failed
// will be moved to primary state and not wait for primaries to be allocated and recovered (*from gateway*)
changed |= electPrimaries(routingNodes);
// now allocate all the unassigned to available nodes
if (routingNodes.hasUnassigned()) {
changed |= allocateUnassigned(routingNodes);
// elect primaries again, in case this is needed with unassigned allocation
changed |= electPrimaries(routingNodes);
}
// elect new primaries (backups that should become primaries)
changed |= electPrimaries(routingNodes);
// rebalance
changed |= rebalance(routingNodes);

View File

@ -0,0 +1,101 @@
/*
* Licensed to Elastic Search and Shay Banon under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. Elastic Search 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.elasticsearch.cluster.routing.strategy;
import org.elasticsearch.cluster.ClusterState;
import org.elasticsearch.cluster.metadata.MetaData;
import org.elasticsearch.cluster.node.DiscoveryNode;
import org.elasticsearch.cluster.routing.RoutingNodes;
import org.elasticsearch.cluster.routing.RoutingTable;
import org.elasticsearch.common.logging.ESLogger;
import org.elasticsearch.common.logging.Loggers;
import org.elasticsearch.common.transport.DummyTransportAddress;
import org.testng.annotations.Test;
import static org.elasticsearch.cluster.ClusterState.*;
import static org.elasticsearch.cluster.metadata.IndexMetaData.*;
import static org.elasticsearch.cluster.metadata.MetaData.*;
import static org.elasticsearch.cluster.node.DiscoveryNodes.*;
import static org.elasticsearch.cluster.routing.RoutingBuilders.*;
import static org.elasticsearch.cluster.routing.ShardRoutingState.*;
import static org.hamcrest.MatcherAssert.*;
import static org.hamcrest.Matchers.*;
/**
* @author kimchy (shay.banon)
*/
public class PrimaryElectionRoutingTests {
private final ESLogger logger = Loggers.getLogger(PrimaryElectionRoutingTests.class);
@Test public void testBackupElectionToPrimaryWhenPrimaryCanBeAllocatedToAnotherNode() {
ShardsRoutingStrategy strategy = new ShardsRoutingStrategy();
logger.info("Building initial routing table");
MetaData metaData = newMetaDataBuilder()
.put(newIndexMetaDataBuilder("test").numberOfShards(1).numberOfReplicas(1))
.build();
RoutingTable routingTable = routingTable()
.add(indexRoutingTable("test").initializeEmpty(metaData.index("test")))
.build();
ClusterState clusterState = newClusterStateBuilder().metaData(metaData).routingTable(routingTable).build();
logger.info("Adding three nodes and performing rerouting");
clusterState = newClusterStateBuilder().state(clusterState).nodes(newNodesBuilder().put(newNode("node1")).put(newNode("node2"))).build();
RoutingTable prevRoutingTable = routingTable;
routingTable = strategy.reroute(clusterState);
clusterState = newClusterStateBuilder().state(clusterState).routingTable(routingTable).build();
logger.info("Start the primary shard (on node1)");
RoutingNodes routingNodes = routingTable.routingNodes(clusterState.metaData());
prevRoutingTable = routingTable;
routingTable = strategy.applyStartedShards(clusterState, routingNodes.node("node1").shardsWithState(INITIALIZING));
clusterState = newClusterStateBuilder().state(clusterState).routingTable(routingTable).build();
logger.info("Start the backup shard (on node2)");
routingNodes = routingTable.routingNodes(clusterState.metaData());
prevRoutingTable = routingTable;
routingTable = strategy.applyStartedShards(clusterState, routingNodes.node("node2").shardsWithState(INITIALIZING));
clusterState = newClusterStateBuilder().state(clusterState).routingTable(routingTable).build();
logger.info("Adding third node and reroute and kill first node");
clusterState = newClusterStateBuilder().state(clusterState).nodes(newNodesBuilder().putAll(clusterState.nodes()).put(newNode("node3")).remove("node1")).build();
prevRoutingTable = routingTable;
routingTable = strategy.reroute(clusterState);
clusterState = newClusterStateBuilder().state(clusterState).routingTable(routingTable).build();
routingNodes = routingTable.routingNodes(metaData);
assertThat(prevRoutingTable != routingTable, equalTo(true));
assertThat(routingTable.index("test").shards().size(), equalTo(1));
assertThat(routingNodes.node("node1"), nullValue());
assertThat(routingNodes.node("node2").numberOfShardsWithState(STARTED), equalTo(1));
assertThat(routingNodes.node("node3").numberOfShardsWithState(INITIALIZING), equalTo(1));
// verify where the primary is
assertThat(routingTable.index("test").shard(0).primaryShard().currentNodeId(), equalTo("node2"));
assertThat(routingTable.index("test").shard(0).backupsShards().get(0).currentNodeId(), equalTo("node3"));
}
private DiscoveryNode newNode(String nodeId) {
return new DiscoveryNode(nodeId, DummyTransportAddress.INSTANCE);
}
}