Add metrics around the file extensions we request when populating the cache (#123134)

This adds the file extentions for the blobs we request when populating the
cache.
The possible values for lucene extensions are around 50 and we use a special
"other" category for everything else, as a fallback.

Author: andreidan

x-pack/plugin/blob-cache/src/main/java/org/elasticsearch/blobcache/BlobCacheMetrics.java

@@ -9,6 +9,7 @@
 
 import org.apache.logging.log4j.LogManager;
 import org.apache.logging.log4j.Logger;
+import org.elasticsearch.index.store.LuceneFilesExtensions;
 import org.elasticsearch.telemetry.TelemetryProvider;
 import org.elasticsearch.telemetry.metric.DoubleHistogram;
 import org.elasticsearch.telemetry.metric.LongCounter;
@@ -24,8 +25,8 @@ public class BlobCacheMetrics {
     private static final double BYTES_PER_NANOSECONDS_TO_MEBIBYTES_PER_SECOND = 1e9D / (1 << 20);
     public static final String CACHE_POPULATION_REASON_ATTRIBUTE_KEY = "reason";
     public static final String CACHE_POPULATION_SOURCE_ATTRIBUTE_KEY = "source";
-    public static final String SHARD_ID_ATTRIBUTE_KEY = "shard_id";
-    public static final String INDEX_ATTRIBUTE_KEY = "index_name";
+    public static final String LUCENE_FILE_EXTENSION_ATTRIBUTE_KEY = "file_extension";
+    public static final String NON_LUCENE_EXTENSION_TO_RECORD = "other";
 
     private final LongCounter cacheMissCounter;
     private final LongCounter evictedCountNonZeroFrequency;
@@ -113,22 +114,28 @@ public LongHistogram getCacheMissLoadTimes() {
     /**
      * Record the various cache population metrics after a chunk is copied to the cache
      *
+     * @param blobName The file that was requested and triggered the cache population.
      * @param bytesCopied The number of bytes copied
      * @param copyTimeNanos The time taken to copy the bytes in nanoseconds
      * @param cachePopulationReason The reason for the cache being populated
      * @param cachePopulationSource The source from which the data is being loaded
      */
     public void recordCachePopulationMetrics(
+        String blobName,
         int bytesCopied,
         long copyTimeNanos,
         CachePopulationReason cachePopulationReason,
         CachePopulationSource cachePopulationSource
     ) {
+        LuceneFilesExtensions luceneFilesExtensions = LuceneFilesExtensions.fromFile(blobName);
+        String blobFileExtension = luceneFilesExtensions != null ? luceneFilesExtensions.getExtension() : NON_LUCENE_EXTENSION_TO_RECORD;
         Map<String, Object> metricAttributes = Map.of(
             CACHE_POPULATION_REASON_ATTRIBUTE_KEY,
             cachePopulationReason.name(),
             CACHE_POPULATION_SOURCE_ATTRIBUTE_KEY,
-            cachePopulationSource.name()
+            cachePopulationSource.name(),
+            LUCENE_FILE_EXTENSION_ATTRIBUTE_KEY,
+            blobFileExtension
         );
         assert bytesCopied > 0 : "We shouldn't be recording zero-sized copies";
         cachePopulationBytes.incrementBy(bytesCopied, metricAttributes);

x-pack/plugin/blob-cache/src/test/java/org/elasticsearch/blobcache/BlobCacheMetricsTests.java

@@ -8,14 +8,18 @@
 package org.elasticsearch.blobcache;
 
 import org.elasticsearch.common.unit.ByteSizeValue;
+import org.elasticsearch.index.store.LuceneFilesExtensions;
 import org.elasticsearch.telemetry.InstrumentType;
 import org.elasticsearch.telemetry.Measurement;
 import org.elasticsearch.telemetry.RecordingMeterRegistry;
 import org.elasticsearch.test.ESTestCase;
 import org.junit.Before;
 
+import java.util.Arrays;
 import java.util.concurrent.TimeUnit;
 
+import static org.hamcrest.Matchers.is;
+
 public class BlobCacheMetricsTests extends ESTestCase {
 
     private RecordingMeterRegistry recordingMeterRegistry;
@@ -32,7 +36,10 @@ public void testRecordCachePopulationMetricsRecordsThroughput() {
         int secondsTaken = randomIntBetween(1, 5);
         BlobCacheMetrics.CachePopulationReason cachePopulationReason = randomFrom(BlobCacheMetrics.CachePopulationReason.values());
         CachePopulationSource cachePopulationSource = randomFrom(CachePopulationSource.values());
+        String fileExtension = randomFrom(Arrays.stream(LuceneFilesExtensions.values()).map(LuceneFilesExtensions::getExtension).toList());
+        String luceneBlobFile = randomAlphanumericOfLength(15) + "." + fileExtension;
         metrics.recordCachePopulationMetrics(
+            luceneBlobFile,
             Math.toIntExact(ByteSizeValue.ofMb(mebiBytesSent).getBytes()),
             TimeUnit.SECONDS.toNanos(secondsTaken),
             cachePopulationReason,
@@ -44,29 +51,31 @@ public void testRecordCachePopulationMetricsRecordsThroughput() {
             .getMeasurements(InstrumentType.DOUBLE_HISTOGRAM, "es.blob_cache.population.throughput.histogram")
             .get(0);
         assertEquals(throughputMeasurement.getDouble(), (double) mebiBytesSent / secondsTaken, 0.0);
-        assertExpectedAttributesPresent(throughputMeasurement, cachePopulationReason, cachePopulationSource);
+        assertExpectedAttributesPresent(throughputMeasurement, cachePopulationReason, cachePopulationSource, fileExtension);
 
         // bytes counter
         Measurement totalBytesMeasurement = recordingMeterRegistry.getRecorder()
             .getMeasurements(InstrumentType.LONG_COUNTER, "es.blob_cache.population.bytes.total")
             .get(0);
         assertEquals(totalBytesMeasurement.getLong(), ByteSizeValue.ofMb(mebiBytesSent).getBytes());
-        assertExpectedAttributesPresent(totalBytesMeasurement, cachePopulationReason, cachePopulationSource);
+        assertExpectedAttributesPresent(totalBytesMeasurement, cachePopulationReason, cachePopulationSource, fileExtension);
 
         // time counter
         Measurement totalTimeMeasurement = recordingMeterRegistry.getRecorder()
             .getMeasurements(InstrumentType.LONG_COUNTER, "es.blob_cache.population.time.total")
             .get(0);
         assertEquals(totalTimeMeasurement.getLong(), TimeUnit.SECONDS.toMillis(secondsTaken));
-        assertExpectedAttributesPresent(totalTimeMeasurement, cachePopulationReason, cachePopulationSource);
+        assertExpectedAttributesPresent(totalTimeMeasurement, cachePopulationReason, cachePopulationSource, fileExtension);
     }
 
     private static void assertExpectedAttributesPresent(
         Measurement measurement,
         BlobCacheMetrics.CachePopulationReason cachePopulationReason,
-        CachePopulationSource cachePopulationSource
+        CachePopulationSource cachePopulationSource,
+        String fileExtension
     ) {
-        assertEquals(measurement.attributes().get(BlobCacheMetrics.CACHE_POPULATION_REASON_ATTRIBUTE_KEY), cachePopulationReason.name());
-        assertEquals(measurement.attributes().get(BlobCacheMetrics.CACHE_POPULATION_SOURCE_ATTRIBUTE_KEY), cachePopulationSource.name());
+        assertThat(measurement.attributes().get(BlobCacheMetrics.CACHE_POPULATION_REASON_ATTRIBUTE_KEY), is(cachePopulationReason.name()));
+        assertThat(measurement.attributes().get(BlobCacheMetrics.CACHE_POPULATION_SOURCE_ATTRIBUTE_KEY), is(cachePopulationSource.name()));
+        assertThat(measurement.attributes().get(BlobCacheMetrics.LUCENE_FILE_EXTENSION_ATTRIBUTE_KEY), is(fileExtension));
     }
 }