Remove zstd feature flag for index codec best compression.

build-tools-internal/src/main/resources/changelog-schema.json

@@ -32,6 +32,7 @@
             "CRUD",
             "Client",
             "Cluster Coordination",
+            "Codec",
             "Data streams",
             "DLM",
             "Discovery-Plugins",

docs/changelog/112665.yaml

@@ -0,0 +1,13 @@
+pr: 112665
+summary: Remove zstd feature flag for index codec best compression
+area: Codec
+type: enhancement
+issues: []
+highlight:
+  title: Enable ZStandard compression for indices with index.codec set to best_compression
+  body: |-
+    Before DEFLATE compression was used to compress stored fields in indices with index.codec index setting set to
+    best_compression, with this change ZStandard is used as compression algorithm to stored fields for indices with
+    index.codec index setting set to best_compression. Experiments have shown that ZStandard offers lower storage usage
+    (upto ~12%) and higher indexing throughput (upto 14%) compared to DEFLATE.
+  notable: true

docs/reference/ilm/actions/ilm-forcemerge.asciidoc

@@ -49,7 +49,7 @@ Number of segments to merge to. To fully merge the index, set to `1`.
 `index_codec`::
 (Optional, string)
 Codec used to compress the document store. The only accepted value is
-`best_compression`, which uses {wikipedia}/DEFLATE[DEFLATE] for a higher
+`best_compression`, which uses {wikipedia}/Zstd[ZSTD] for a higher
 compression ratio but slower stored fields performance. To use the default LZ4
 codec, omit this argument.
 +

docs/reference/index-modules.asciidoc

@@ -76,14 +76,16 @@ breaking change].
 
     The +default+ value compresses stored data with LZ4
     compression, but this can be set to +best_compression+
-    which uses {wikipedia}/DEFLATE[DEFLATE] for a higher
-    compression ratio, at the expense of slower stored fields performance.
+    which uses {wikipedia}/Zstd[ZSTD] for a higher
+    compression ratio, at the expense of slower stored fields read performance.
     If you are updating the compression type, the new one will be applied
     after segments are merged. Segment merging can be forced using
     <<indices-forcemerge,force merge>>. Experiments with indexing log datasets
-    have shown that `best_compression` gives up to ~18% lower storage usage in
-    the most ideal scenario compared to `default` while only minimally affecting
-    indexing throughput (~2%).
+    have shown that `best_compression` gives up to ~28% lower storage usage and
+    similar indexing throughput (sometimes a bit slower or faster depending on other used options) compared
+    to `default` while affecting get by id latencies between ~10% and ~33%. The higher get
+    by id latencies is not a concern for many use cases like logging or metrics, since
+    these don't really rely on get by id functionality (Get APIs or searching by _id).
 
 [[index-mode-setting]] `index.mode`::
 +

server/src/main/java/org/elasticsearch/index/codec/CodecService.java

@@ -53,15 +53,11 @@ public CodecService(@Nullable MapperService mapperService, BigArrays bigArrays)
         }
         codecs.put(LEGACY_DEFAULT_CODEC, legacyBestSpeedCodec);
 
+        codecs.put(
+            BEST_COMPRESSION_CODEC,
+            new PerFieldMapperCodec(Zstd814StoredFieldsFormat.Mode.BEST_COMPRESSION, mapperService, bigArrays)
+        );
         Codec legacyBestCompressionCodec = new LegacyPerFieldMapperCodec(Lucene99Codec.Mode.BEST_COMPRESSION, mapperService, bigArrays);
-        if (ZSTD_STORED_FIELDS_FEATURE_FLAG.isEnabled()) {
-            codecs.put(
-                BEST_COMPRESSION_CODEC,
-                new PerFieldMapperCodec(Zstd814StoredFieldsFormat.Mode.BEST_COMPRESSION, mapperService, bigArrays)
-            );
-        } else {
-            codecs.put(BEST_COMPRESSION_CODEC, legacyBestCompressionCodec);
-        }
         codecs.put(LEGACY_BEST_COMPRESSION_CODEC, legacyBestCompressionCodec);
 
         codecs.put(LUCENE_DEFAULT_CODEC, Codec.getDefault());

server/src/test/java/org/elasticsearch/index/codec/CodecIntegrationTests.java

@@ -17,8 +17,6 @@
 public class CodecIntegrationTests extends ESSingleNodeTestCase {
 
     public void testCanConfigureLegacySettings() {
-        assumeTrue("Only when zstd_stored_fields feature flag is enabled", CodecService.ZSTD_STORED_FIELDS_FEATURE_FLAG.isEnabled());
-
         createIndex("index1", Settings.builder().put("index.codec", "legacy_default").build());
         var codec = client().admin().indices().prepareGetSettings("index1").execute().actionGet().getSetting("index1", "index.codec");
         assertThat(codec, equalTo("legacy_default"));
@@ -29,8 +27,6 @@ public void testCanConfigureLegacySettings() {
     }
 
     public void testDefaultCodecLogsdb() {
-        assumeTrue("Only when zstd_stored_fields feature flag is enabled", CodecService.ZSTD_STORED_FIELDS_FEATURE_FLAG.isEnabled());
-
         var indexService = createIndex("index1", Settings.builder().put("index.mode", "logsdb").build());
         var storedFieldsFormat = (Zstd814StoredFieldsFormat) indexService.getShard(0)
             .getEngineOrNull()

server/src/test/java/org/elasticsearch/index/codec/CodecTests.java

@@ -64,7 +64,6 @@ public void testDefault() throws Exception {
     }
 
     public void testBestCompression() throws Exception {
-        assumeTrue("Only when zstd_stored_fields feature flag is enabled", CodecService.ZSTD_STORED_FIELDS_FEATURE_FLAG.isEnabled());
         Codec codec = createCodecService().codec("best_compression");
         assertEquals(
             "Zstd814StoredFieldsFormat(compressionMode=ZSTD(level=3), chunkSize=245760, maxDocsPerChunk=2048, blockShift=10)",