Fix test, and update javadoc

thecoop · Jan 8, 2025 · 00f52b9 · 00f52b9
1 parent 665222b
commit 00f52b9
Show file tree

Hide file tree

Showing 2 changed files with 12 additions and 2 deletions.
diff --git a/server/src/main/java/org/elasticsearch/features/FeatureSpecification.java b/server/src/main/java/org/elasticsearch/features/FeatureSpecification.java
@@ -20,8 +20,14 @@
  * unless they also support that feature (this is known as the 'feature ratchet').
  * So once a feature is supported by a cluster, it will always be supported by that cluster in the future.
  * <p>
+ * The only situation where features can be removed is on a compatibility boundary (normally a new major version).
+ * Because a new major version can only form a cluster with the highest minor of the previous major,
+ * any features introduced before that point can be assumed to always be present on that cluster.
+ * The feature itself can be removed by marking it as assumed by {@link NodeFeature#assumedAfterNextCompatibilityBoundary()},
+ * in the last minor, and then removing it and associated conditions in the new major version.
+ * <p>
  * The feature information in cluster state should not normally be directly accessed.
- * All feature checks should be done through {@code FeatureService} to ensure that Elasticsearch's
+ * All feature checks should be done through {@link FeatureService} to ensure that Elasticsearch's
  * guarantees on the introduction of new functionality are followed;
  * that is, new functionality is not enabled until all nodes in the cluster support it.
  * <p>

diff --git a/...org/elasticsearch/action/datastreams/autosharding/DataStreamAutoShardingServiceTests.java b/...org/elasticsearch/action/datastreams/autosharding/DataStreamAutoShardingServiceTests.java
@@ -144,7 +144,11 @@ public Set<NodeFeature> getFeatures() {
 
         {
             // cluster doesn't have feature
-            ClusterState stateNoFeature = ClusterState.builder(ClusterName.DEFAULT).metadata(Metadata.builder()).build();
+            ClusterState stateNoFeature = ClusterState.builder(ClusterName.DEFAULT)
+                .nodes(DiscoveryNodes.builder().add(DiscoveryNodeUtils.create("n1")).add(DiscoveryNodeUtils.create("n2")))
+                .nodeFeatures(Map.of("n1", Set.of(), "n2", Set.of()))
+                .metadata(Metadata.builder())
+                .build();
 
             Settings settings = Settings.builder().put(DataStreamAutoShardingService.DATA_STREAMS_AUTO_SHARDING_ENABLED, true).build();
             DataStreamAutoShardingService noFeatureService = new DataStreamAutoShardingService(