/*
* Licensed to Elasticsearch under one or more contributor
* license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright
* ownership. Elasticsearch 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.allocation.decider;
import com.carrotsearch.hppc.ObjectIntHashMap;
import com.google.common.collect.Maps;
import org.elasticsearch.cluster.metadata.IndexMetaData;
import org.elasticsearch.cluster.routing.RoutingNode;
import org.elasticsearch.cluster.routing.ShardRouting;
import org.elasticsearch.cluster.routing.allocation.RoutingAllocation;
import org.elasticsearch.common.Strings;
import org.elasticsearch.common.inject.Inject;
import org.elasticsearch.common.settings.Settings;
import org.elasticsearch.node.settings.NodeSettingsService;
import java.util.HashMap;
import java.util.Map;
/**
* This {@link AllocationDecider} controls shard allocation based on
* <tt>awareness</tt> key-value pairs defined in the node configuration.
* Awareness explicitly controls where replicas should be allocated based on
* attributes like node or physical rack locations. Awareness attributes accept
* arbitrary configuration keys like a rack data-center identifier. For example
* the setting:
* <pre>
* cluster.routing.allocation.awareness.attributes: rack_id
* </pre>
* <p>
* will cause allocations to be distributed over different racks such that
* ideally at least one replicas of the all shard is available on the same rack.
* To enable allocation awareness in this example nodes should contain a value
* for the <tt>rack_id</tt> key like:
* <pre>
* node.rack_id:1
* </pre>
* <p>
* Awareness can also be used to prevent over-allocation in the case of node or
* even "zone" failure. For example in cloud-computing infrastructures like
* Amazone AWS a cluster might span over multiple "zones". Awareness can be used
* to distribute replicas to individual zones by setting:
* <pre>
* cluster.routing.allocation.awareness.attributes: zone
* </pre>
* <p>
* and forcing allocation to be aware of the following zone the data resides in:
* <pre>
* cluster.routing.allocation.awareness.force.zone.values: zone1,zone2
* </pre>
* <p>
* In contrast to regular awareness this setting will prevent over-allocation on
* <tt>zone1</tt> even if <tt>zone2</tt> fails partially or becomes entirely
* unavailable. Nodes that belong to a certain zone / group should be started
* with the zone id configured on the node-level settings like:
* <pre>
* node.zone: zone1
* </pre>
*/
public class AwarenessAllocationDecider extends AllocationDecider {
public static final String NAME = "awareness";
public static final String CLUSTER_ROUTING_ALLOCATION_AWARENESS_ATTRIBUTES = "cluster.routing.allocation.awareness.attributes";
public static final String CLUSTER_ROUTING_ALLOCATION_AWARENESS_FORCE_GROUP = "cluster.routing.allocation.awareness.force.";
class ApplySettings implements NodeSettingsService.Listener {
@Override
public void onRefreshSettings(Settings settings) {
String[] awarenessAttributes = settings.getAsArray(CLUSTER_ROUTING_ALLOCATION_AWARENESS_ATTRIBUTES, null);
if (awarenessAttributes == null && "".equals(settings.get(CLUSTER_ROUTING_ALLOCATION_AWARENESS_ATTRIBUTES, null))) {
awarenessAttributes = Strings.EMPTY_ARRAY; // the empty string resets this
}
if (awarenessAttributes != null) {
logger.info("updating [cluster.routing.allocation.awareness.attributes] from [{}] to [{}]", AwarenessAllocationDecider.this.awarenessAttributes, awarenessAttributes);
AwarenessAllocationDecider.this.awarenessAttributes = awarenessAttributes;
}
Map<String, String[]> forcedAwarenessAttributes = new HashMap<>(AwarenessAllocationDecider.this.forcedAwarenessAttributes);
Map<String, Settings> forceGroups = settings.getGroups(CLUSTER_ROUTING_ALLOCATION_AWARENESS_FORCE_GROUP);
if (!forceGroups.isEmpty()) {
for (Map.Entry<String, Settings> entry : forceGroups.entrySet()) {
String[] aValues = entry.getValue().getAsArray("values");
if (aValues.length > 0) {
forcedAwarenessAttributes.put(entry.getKey(), aValues);
}
}
}
AwarenessAllocationDecider.this.forcedAwarenessAttributes = forcedAwarenessAttributes;
}
}
private String[] awarenessAttributes;
private Map<String, String[]> forcedAwarenessAttributes;
/**
* Creates a new {@link AwarenessAllocationDecider} instance
*/
public AwarenessAllocationDecider() {
this(Settings.Builder.EMPTY_SETTINGS);
}
/**
* Creates a new {@link AwarenessAllocationDecider} instance from given settings
*
* @param settings {@link Settings} to use
*/
public AwarenessAllocationDecider(Settings settings) {
this(settings, new NodeSettingsService(settings));
}
@Inject
public AwarenessAllocationDecider(Settings settings, NodeSettingsService nodeSettingsService) {
super(settings);
this.awarenessAttributes = settings.getAsArray(CLUSTER_ROUTING_ALLOCATION_AWARENESS_ATTRIBUTES);
forcedAwarenessAttributes = Maps.newHashMap();
Map<String, Settings> forceGroups = settings.getGroups(CLUSTER_ROUTING_ALLOCATION_AWARENESS_FORCE_GROUP);
for (Map.Entry<String, Settings> entry : forceGroups.entrySet()) {
String[] aValues = entry.getValue().getAsArray("values");
if (aValues.length > 0) {
forcedAwarenessAttributes.put(entry.getKey(), aValues);
}
}
nodeSettingsService.addListener(new ApplySettings());
}
/**
* Get the attributes defined by this instance
*
* @return attributes defined by this instance
*/
public String[] awarenessAttributes() {
return this.awarenessAttributes;
}
@Override
public Decision canAllocate(ShardRouting shardRouting, RoutingNode node, RoutingAllocation allocation) {
return underCapacity(shardRouting, node, allocation, true);
}
@Override
public Decision canRemain(ShardRouting shardRouting, RoutingNode node, RoutingAllocation allocation) {
return underCapacity(shardRouting, node, allocation, false);
}
private Decision underCapacity(ShardRouting shardRouting, RoutingNode node, RoutingAllocation allocation, boolean moveToNode) {
if (awarenessAttributes.length == 0) {
return allocation.decision(Decision.YES, NAME, "no allocation awareness enabled");
}
IndexMetaData indexMetaData = allocation.metaData().index(shardRouting.index());
int shardCount = indexMetaData.getNumberOfReplicas() + 1; // 1 for primary
for (String awarenessAttribute : awarenessAttributes) {
// the node the shard exists on must be associated with an awareness attribute
if (!node.node().attributes().containsKey(awarenessAttribute)) {
return allocation.decision(Decision.NO, NAME, "node does not contain awareness attribute: [%s]", awarenessAttribute);
}
// build attr_value -> nodes map
ObjectIntHashMap<String> nodesPerAttribute = allocation.routingNodes().nodesPerAttributesCounts(awarenessAttribute);
// build the count of shards per attribute value
ObjectIntHashMap<String> shardPerAttribute = new ObjectIntHashMap<>();
for (ShardRouting assignedShard : allocation.routingNodes().assignedShards(shardRouting)) {
if (assignedShard.started() || assignedShard.initializing()) {
// Note: this also counts relocation targets as that will be the new location of the shard.
// Relocation sources should not be counted as the shard is moving away
RoutingNode routingNode = allocation.routingNodes().node(assignedShard.currentNodeId());
shardPerAttribute.addTo(routingNode.node().attributes().get(awarenessAttribute), 1);
}
}
if (moveToNode) {
if (shardRouting.assignedToNode()) {
String nodeId = shardRouting.relocating() ? shardRouting.relocatingNodeId() : shardRouting.currentNodeId();
if (!node.nodeId().equals(nodeId)) {
// we work on different nodes, move counts around
shardPerAttribute.putOrAdd(allocation.routingNodes().node(nodeId).node().attributes().get(awarenessAttribute), 0, -1);
shardPerAttribute.addTo(node.node().attributes().get(awarenessAttribute), 1);
}
} else {
shardPerAttribute.addTo(node.node().attributes().get(awarenessAttribute), 1);
}
}
int numberOfAttributes = nodesPerAttribute.size();
String[] fullValues = forcedAwarenessAttributes.get(awarenessAttribute);
if (fullValues != null) {
for (String fullValue : fullValues) {
if (!shardPerAttribute.containsKey(fullValue)) {
numberOfAttributes++;
}
}
}
// TODO should we remove ones that are not part of full list?
int averagePerAttribute = shardCount / numberOfAttributes;
int totalLeftover = shardCount % numberOfAttributes;
int requiredCountPerAttribute;
if (averagePerAttribute == 0) {
// if we have more attributes values than shard count, no leftover
totalLeftover = 0;
requiredCountPerAttribute = 1;
} else {
requiredCountPerAttribute = averagePerAttribute;
}
int leftoverPerAttribute = totalLeftover == 0 ? 0 : 1;
int currentNodeCount = shardPerAttribute.get(node.node().attributes().get(awarenessAttribute));
// if we are above with leftover, then we know we are not good, even with mod
if (currentNodeCount > (requiredCountPerAttribute + leftoverPerAttribute)) {
return allocation.decision(Decision.NO, NAME,
"too many shards on node for attribute: [%s], required per attribute: [%d], node count: [%d], leftover: [%d]",
awarenessAttribute, requiredCountPerAttribute, currentNodeCount, leftoverPerAttribute);
}
// all is well, we are below or same as average
if (currentNodeCount <= requiredCountPerAttribute) {
continue;
}
}
return allocation.decision(Decision.YES, NAME, "node meets awareness requirements");
}
}