honeymoose
/
OpenSearch
mirror of https://github.com/honeymoose/OpenSearch.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Skipping execution of remaining Deciders if one of them returns a Decision.NO. 

Reordering execution of the different AllocationDeciders, so the cheapest get executed first.
This commit is contained in:
Sebastian Geidies 2013-11-26 11:00:46 +01:00 committed by Simon Willnauer
parent 2acccf257d
commit 6ed126deaf
1 changed files with 48 additions and 14 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

58
src/main/java/org/elasticsearch/cluster/routing/allocation/decider/AllocationDeciders.java
View File

@ -38,25 +38,50 @@ public class AllocationDeciders extends AllocationDecider {
private final AllocationDecider[] allocations;
/**
* Create a new {@link AllocationDeciders} instance
* Create a new {@link AllocationDeciders} instance. The different deciders
* should be added in order, as looping over them will stop when the first
* return a {@link Decision.THROTTLE} or {@link Decision.NO}. For performance
* reasons, those more likely to return either of these, and those with
* cheap execution should be executed first.
*
* Performance characteristics:
* {@link ConcurrentRebalanceAllocationDecider} numerical comparison of a counter in {@link RoutingNodes},
* constant performance, likely to be triggered.
* {@link DisableAllocationDecider} lookup of setting. Constant performance, not as
* likely to be triggered.
* {@link ClusterRebalanceAllocationDecider} checks for unassigned primaries, inactive primaries and
* a rebalance already happening in replica set.
* {@link DiskThresholdDecider} one numerical comparison per node in cluster.
* {@link SnapshotInProgressAllocationDecider} status lookup, unlikely.
* {@link FilterAllocationDecider} checks all allocation include/exclude filters in the cluster against the
* node's attributes.
* {@link RebalanceOnlyWhenActiveAllocationDecider} checks if all shards are active.
* {@link ReplicaAfterPrimaryActiveAllocationDecider} finds primary in replica set, checks whether it
* is started.
* {@link ShardsLimitAllocationDecider} loops over shards allocated on a node, filters out non-relocating
* shards of the same index to do a count comparison.
* {@link AwarenessAllocationDecider} loops over all shards in cluster.
* {@link SameShardAllocationDecider} loops over shards on node.
* {@link ThrottlingAllocationDecider} checks primaries initializing (looping over shards on node) for a primary
* to be allocated, for replicas loops over all shards on node.
*
* @param settings settings to use
* @param nodeSettingsService per-node settings to use
*/
public AllocationDeciders(Settings settings, NodeSettingsService nodeSettingsService) {
this(settings, ImmutableSet.<AllocationDecider>builder()
.add(new SameShardAllocationDecider(settings))
.add(new FilterAllocationDecider(settings, nodeSettingsService))
.add(new ReplicaAfterPrimaryActiveAllocationDecider(settings))
.add(new ThrottlingAllocationDecider(settings, nodeSettingsService))
.add(new RebalanceOnlyWhenActiveAllocationDecider(settings))
.add(new ClusterRebalanceAllocationDecider(settings))
.add(new ConcurrentRebalanceAllocationDecider(settings, nodeSettingsService))
.add(new DisableAllocationDecider(settings, nodeSettingsService))
.add(new AwarenessAllocationDecider(settings, nodeSettingsService))
.add(new ShardsLimitAllocationDecider(settings))
.add(new ClusterRebalanceAllocationDecider(settings))
.add(new DiskThresholdDecider(settings, nodeSettingsService))
.add(new SnapshotInProgressAllocationDecider(settings))
.add(new FilterAllocationDecider(settings, nodeSettingsService))
.add(new RebalanceOnlyWhenActiveAllocationDecider(settings))
.add(new ReplicaAfterPrimaryActiveAllocationDecider(settings))
.add(new ShardsLimitAllocationDecider(settings))
.add(new AwarenessAllocationDecider(settings, nodeSettingsService))
.add(new SameShardAllocationDecider(settings))
.add(new ThrottlingAllocationDecider(settings, nodeSettingsService))
.build()
);
}
@ -72,7 +97,10 @@ public class AllocationDeciders extends AllocationDecider {
Decision.Multi ret = new Decision.Multi();
for (AllocationDecider allocationDecider : allocations) {
Decision decision = allocationDecider.canRebalance(shardRouting, allocation);
if (decision != Decision.ALWAYS) {
// short track if a NO is returned.
if (decision == Decision.NO) {
return decision;
} else if (decision != Decision.ALWAYS) {
ret.add(decision);
}
}
@ -87,9 +115,12 @@ public class AllocationDeciders extends AllocationDecider {
Decision.Multi ret = new Decision.Multi();
for (AllocationDecider allocationDecider : allocations) {
Decision decision = allocationDecider.canAllocate(shardRouting, node, allocation);
// short track if a NO is returned.
if (decision == Decision.NO) {
return decision;
} else if (decision != Decision.ALWAYS) {
// the assumption is that a decider that returns the static instance Decision#ALWAYS
// does not really implements canAllocate
if (decision != Decision.ALWAYS) {
ret.add(decision);
}
}
@ -104,7 +135,10 @@ public class AllocationDeciders extends AllocationDecider {
Decision.Multi ret = new Decision.Multi();
for (AllocationDecider allocationDecider : allocations) {
Decision decision = allocationDecider.canRemain(shardRouting, node, allocation);
if (decision != Decision.ALWAYS) {
// short track if a NO is returned.
if (decision == Decision.NO) {
return decision;
} else if (decision != Decision.ALWAYS) {
ret.add(decision);
}
}