Migrate peer recovery from translog to retention lease (#49448)

Since 7.4, we switch from translog to Lucene as the source of history
for peer recoveries. However, we reduce the likelihood of
operation-based recoveries when performing a full cluster restart from
pre-7.4 because existing copies do not have PPRL.

To remedy this issue, we fallback using translog in peer recoveries if
the recovering replica does not have a peer recovery retention lease,
and the replication group hasn't fully migrated to PRRL.

Relates #45136

Author: dnhatn

qa/full-cluster-restart/src/test/java/org/elasticsearch/upgrades/FullClusterRestartIT.java

@@ -35,6 +35,7 @@
 import org.elasticsearch.common.xcontent.XContentBuilder;
 import org.elasticsearch.common.xcontent.json.JsonXContent;
 import org.elasticsearch.common.xcontent.support.XContentMapValues;
+import org.elasticsearch.index.IndexSettings;
 import org.elasticsearch.index.seqno.RetentionLeaseUtils;
 import org.elasticsearch.rest.action.document.RestBulkAction;
 import org.elasticsearch.rest.action.document.RestGetAction;
@@ -1267,6 +1268,12 @@ private void indexRandomDocuments(
         }
     }
 
+    private void indexDocument(String id) throws IOException {
+        final Request indexRequest = new Request("POST", "/" + index + "/" + "_doc/" + id);
+        indexRequest.setJsonEntity(Strings.toString(JsonXContent.contentBuilder().startObject().field("f", "v").endObject()));
+        assertOK(client().performRequest(indexRequest));
+    }
+
     private int countOfIndexedRandomDocuments() throws IOException {
         return Integer.parseInt(loadInfoDocument("count"));
     }
@@ -1362,4 +1369,63 @@ public void testPeerRecoveryRetentionLeases() throws IOException {
             RetentionLeaseUtils.assertAllCopiesHavePeerRecoveryRetentionLeases(client(), index);
         }
     }
+
+    /**
+     * Tests that with or without soft-deletes, we should perform an operation-based recovery if there were some
+     * but not too many uncommitted documents (i.e., less than 10% of committed documents or the extra translog)
+     * before we restart the cluster. This is important when we move from translog based to retention leases based
+     * peer recoveries.
+     */
+    public void testOperationBasedRecovery() throws Exception {
+        if (isRunningAgainstOldCluster()) {
+            createIndex(index, Settings.builder()
+                .put(IndexMetaData.SETTING_NUMBER_OF_SHARDS, 1)
+                .put(IndexMetaData.SETTING_NUMBER_OF_REPLICAS, 1)
+                .put(IndexSettings.INDEX_SOFT_DELETES_SETTING.getKey(), randomBoolean())
+                .build());
+            ensureGreen(index);
+            int committedDocs = randomIntBetween(100, 200);
+            for (int i = 0; i < committedDocs; i++) {
+                indexDocument(Integer.toString(i));
+                if (rarely()) {
+                    flush(index, randomBoolean());
+                }
+            }
+            flush(index, true);
+            ensurePeerRecoveryRetentionLeasesRenewedAndSynced(index);
+            // less than 10% of the committed docs (see IndexSetting#FILE_BASED_RECOVERY_THRESHOLD_SETTING).
+            int uncommittedDocs = randomIntBetween(0, (int) (committedDocs * 0.1));
+            for (int i = 0; i < uncommittedDocs; i++) {
+                final String id = Integer.toString(randomIntBetween(1, 100));
+                indexDocument(id);
+            }
+        } else {
+            ensureGreen(index);
+            assertNoFileBasedRecovery(index, n -> true);
+        }
+    }
+
+    /**
+     * Verifies that once all shard copies on the new version, we should turn off the translog retention for indices with soft-deletes.
+     */
+    public void testTurnOffTranslogRetentionAfterUpgraded() throws Exception {
+        if (isRunningAgainstOldCluster()) {
+            createIndex(index, Settings.builder()
+                .put(IndexMetaData.INDEX_NUMBER_OF_SHARDS_SETTING.getKey(), 1)
+                .put(IndexMetaData.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), 1)
+                .put(IndexSettings.INDEX_SOFT_DELETES_SETTING.getKey(), true).build());
+            ensureGreen(index);
+            int numDocs = randomIntBetween(10, 100);
+            for (int i = 0; i < numDocs; i++) {
+                indexDocument(Integer.toString(randomIntBetween(1, 100)));
+                if (rarely()) {
+                    flush(index, randomBoolean());
+                }
+            }
+        } else {
+            ensureGreen(index);
+            flush(index, true);
+            assertEmptyTranslog(index);
+        }
+    }
 }

qa/rolling-upgrade/src/test/java/org/elasticsearch/upgrades/RecoveryIT.java

@@ -516,10 +516,10 @@ public void testClosedIndexNoopRecovery() throws Exception {
                 switch (CLUSTER_TYPE) {
                     case OLD: break;
                     case MIXED:
-                        assertNoFileBasedRecovery(indexName, s -> s.startsWith(CLUSTER_NAME + "-0"));
+                        assertNoopRecoveries(indexName, s -> s.startsWith(CLUSTER_NAME + "-0"));
                         break;
                     case UPGRADED:
-                        assertNoFileBasedRecovery(indexName, s -> s.startsWith(CLUSTER_NAME));
+                        assertNoopRecoveries(indexName, s -> s.startsWith(CLUSTER_NAME));
                         break;
                 }
             }
@@ -692,7 +692,7 @@ public void testUpdateDoc() throws Exception {
         }
     }
 
-    private void assertNoFileBasedRecovery(String indexName, Predicate<String> targetNode) throws IOException {
+    private void assertNoopRecoveries(String indexName, Predicate<String> targetNode) throws IOException {
         Map<String, Object> recoveries = entityAsMap(client()
             .performRequest(new Request("GET", indexName + "/_recovery?detailed=true")));
 
@@ -723,4 +723,55 @@ private void assertNoFileBasedRecovery(String indexName, Predicate<String> targe
 
         assertTrue("must find replica", foundReplica);
     }
+
+    /**
+     * Tests that with or without soft-deletes, we should perform an operation-based recovery if there were some
+     * but not too many uncommitted documents (i.e., less than 10% of committed documents or the extra translog)
+     * before we upgrade each node. This is important when we move from translog based to retention leases based
+     * peer recoveries.
+     */
+    public void testOperationBasedRecovery() throws Exception {
+        final String index = "test_operation_based_recovery";
+        if (CLUSTER_TYPE == ClusterType.OLD) {
+            createIndex(index, Settings.builder()
+                .put(IndexMetaData.INDEX_NUMBER_OF_SHARDS_SETTING.getKey(), 1)
+                .put(IndexMetaData.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), 2)
+                .put(IndexSettings.INDEX_SOFT_DELETES_SETTING.getKey(), randomBoolean()).build());
+            ensureGreen(index);
+            indexDocs(index, 0, randomIntBetween(100, 200));
+            flush(index, randomBoolean());
+            ensurePeerRecoveryRetentionLeasesRenewedAndSynced(index);
+            // uncommitted docs must be less than 10% of committed docs (see IndexSetting#FILE_BASED_RECOVERY_THRESHOLD_SETTING).
+            indexDocs(index, randomIntBetween(0, 100), randomIntBetween(0, 3));
+        } else {
+            ensureGreen(index);
+            assertNoFileBasedRecovery(index, nodeName ->
+                   CLUSTER_TYPE == ClusterType.UPGRADED
+                || nodeName.startsWith(CLUSTER_NAME + "-0")
+                || (nodeName.startsWith(CLUSTER_NAME + "-1") && Booleans.parseBoolean(System.getProperty("tests.first_round")) == false));
+            indexDocs(index, randomIntBetween(0, 100), randomIntBetween(0, 3));
+        }
+    }
+
+    /**
+     * Verifies that once all shard copies on the new version, we should turn off the translog retention for indices with soft-deletes.
+     */
+    public void testTurnOffTranslogRetentionAfterUpgraded() throws Exception {
+        final String index = "turn_off_translog_retention";
+        if (CLUSTER_TYPE == ClusterType.OLD) {
+            createIndex(index, Settings.builder()
+                .put(IndexMetaData.INDEX_NUMBER_OF_SHARDS_SETTING.getKey(), 1)
+                .put(IndexMetaData.INDEX_NUMBER_OF_REPLICAS_SETTING.getKey(), randomIntBetween(0, 2))
+                .put(IndexSettings.INDEX_SOFT_DELETES_SETTING.getKey(), true).build());
+            ensureGreen(index);
+            indexDocs(index, 0, randomIntBetween(100, 200));
+            flush(index, randomBoolean());
+            indexDocs(index, randomIntBetween(0, 100), randomIntBetween(0, 100));
+        }
+        if (CLUSTER_TYPE == ClusterType.UPGRADED) {
+            ensureGreen(index);
+            flush(index, true);
+            assertEmptyTranslog(index);
+        }
+    }
 }

server/src/main/java/org/elasticsearch/index/IndexSettings.java

@@ -252,21 +252,22 @@ public final class IndexSettings {
      * Controls how long translog files that are no longer needed for persistence reasons
      * will be kept around before being deleted. Keeping more files is useful to increase
      * the chance of ops based recoveries for indices with soft-deletes disabled.
-     * This setting will be ignored if soft-deletes is enabled.
+     * This setting will be ignored if soft-deletes is used in peer recoveries (default in 7.4).
      **/
     public static final Setting<TimeValue> INDEX_TRANSLOG_RETENTION_AGE_SETTING =
         Setting.timeSetting("index.translog.retention.age",
-            settings -> INDEX_SOFT_DELETES_SETTING.get(settings) ? TimeValue.MINUS_ONE : TimeValue.timeValueHours(12), TimeValue.MINUS_ONE,
-            Property.Dynamic, Property.IndexScope);
+            settings -> shouldDisableTranslogRetention(settings) ? TimeValue.MINUS_ONE : TimeValue.timeValueHours(12),
+            TimeValue.MINUS_ONE, Property.Dynamic, Property.IndexScope);
 
     /**
      * Controls how many translog files that are no longer needed for persistence reasons
      * will be kept around before being deleted. Keeping more files is useful to increase
      * the chance of ops based recoveries for indices with soft-deletes disabled.
-     * This setting will be ignored if soft-deletes is enabled.
+     * This setting will be ignored if soft-deletes is used in peer recoveries (default in 7.4).
      **/
     public static final Setting<ByteSizeValue> INDEX_TRANSLOG_RETENTION_SIZE_SETTING =
-        Setting.byteSizeSetting("index.translog.retention.size", settings -> INDEX_SOFT_DELETES_SETTING.get(settings) ? "-1" : "512MB",
+        Setting.byteSizeSetting("index.translog.retention.size",
+            settings -> shouldDisableTranslogRetention(settings) ? "-1" : "512MB",
             Property.Dynamic, Property.IndexScope);
 
     /**
@@ -587,7 +588,7 @@ private void setFlushAfterMergeThresholdSize(ByteSizeValue byteSizeValue) {
     }
 
     private void setTranslogRetentionSize(ByteSizeValue byteSizeValue) {
-        if (softDeleteEnabled && byteSizeValue.getBytes() >= 0) {
+        if (shouldDisableTranslogRetention(settings) && byteSizeValue.getBytes() >= 0) {
             // ignore the translog retention settings if soft-deletes enabled
             this.translogRetentionSize = new ByteSizeValue(-1);
         } else {
@@ -596,7 +597,7 @@ private void setTranslogRetentionSize(ByteSizeValue byteSizeValue) {
     }
 
     private void setTranslogRetentionAge(TimeValue age) {
-        if (softDeleteEnabled && age.millis() >= 0) {
+        if (shouldDisableTranslogRetention(settings) && age.millis() >= 0) {
             // ignore the translog retention settings if soft-deletes enabled
             this.translogRetentionAge = TimeValue.MINUS_ONE;
         } else {
@@ -784,7 +785,7 @@ public TimeValue getRefreshInterval() {
      * Returns the transaction log retention size which controls how much of the translog is kept around to allow for ops based recoveries
      */
     public ByteSizeValue getTranslogRetentionSize() {
-        assert softDeleteEnabled == false || translogRetentionSize.getBytes() == -1L : translogRetentionSize;
+        assert shouldDisableTranslogRetention(settings) == false || translogRetentionSize.getBytes() == -1L : translogRetentionSize;
         return translogRetentionSize;
     }
 
@@ -793,7 +794,7 @@ public ByteSizeValue getTranslogRetentionSize() {
      * around
      */
     public TimeValue getTranslogRetentionAge() {
-        assert softDeleteEnabled == false || translogRetentionAge.millis() == -1L : translogRetentionSize;
+        assert shouldDisableTranslogRetention(settings) == false || translogRetentionAge.millis() == -1L : translogRetentionSize;
         return translogRetentionAge;
     }
 
@@ -805,6 +806,11 @@ public int getTranslogRetentionTotalFiles() {
         return INDEX_TRANSLOG_RETENTION_TOTAL_FILES_SETTING.get(getSettings());
     }
 
+    private static boolean shouldDisableTranslogRetention(Settings settings) {
+        return INDEX_SOFT_DELETES_SETTING.get(settings)
+            && IndexMetaData.SETTING_INDEX_VERSION_CREATED.get(settings).onOrAfter(Version.V_7_4_0);
+    }
+
     /**
      * Returns the generation threshold size. As sequence numbers can cause multiple generations to
      * be preserved for rollback purposes, we want to keep the size of individual generations from

server/src/main/java/org/elasticsearch/index/engine/Engine.java

@@ -66,6 +66,7 @@
 import org.elasticsearch.common.lucene.uid.VersionsAndSeqNoResolver.DocIdAndVersion;
 import org.elasticsearch.common.metrics.CounterMetric;
 import org.elasticsearch.common.regex.Regex;
+import org.elasticsearch.common.unit.ByteSizeValue;
 import org.elasticsearch.common.unit.TimeValue;
 import org.elasticsearch.common.util.concurrent.ReleasableLock;
 import org.elasticsearch.index.VersionType;
@@ -729,7 +730,7 @@ public enum SearcherScope {
     /**
      * Acquires a lock on the translog files and Lucene soft-deleted documents to prevent them from being trimmed
      */
-    public abstract Closeable acquireRetentionLock();
+    public abstract Closeable acquireHistoryRetentionLock(HistorySource historySource);
 
     /**
      * Creates a new history snapshot from Lucene for reading operations whose seqno in the requesting seqno range (both inclusive).
@@ -742,19 +743,20 @@ public abstract Translog.Snapshot newChangesSnapshot(String source, MapperServic
      * Creates a new history snapshot for reading operations since {@code startingSeqNo} (inclusive).
      * The returned snapshot can be retrieved from either Lucene index or translog files.
      */
-    public abstract Translog.Snapshot readHistoryOperations(String source,
-                                                                MapperService mapperService, long startingSeqNo) throws IOException;
+    public abstract Translog.Snapshot readHistoryOperations(String reason, HistorySource historySource,
+                                                            MapperService mapperService, long startingSeqNo) throws IOException;
 
     /**
      * Returns the estimated number of history operations whose seq# at least {@code startingSeqNo}(inclusive) in this engine.
      */
-    public abstract int estimateNumberOfHistoryOperations(String source,
-                                                                MapperService mapperService, long startingSeqNo) throws IOException;
+    public abstract int estimateNumberOfHistoryOperations(String reason, HistorySource historySource,
+                                                          MapperService mapperService, long startingSeqNo) throws IOException;
 
     /**
      * Checks if this engine has every operations since  {@code startingSeqNo}(inclusive) in its history (either Lucene or translog)
      */
-    public abstract boolean hasCompleteOperationHistory(String source, MapperService mapperService, long startingSeqNo) throws IOException;
+    public abstract boolean hasCompleteOperationHistory(String reason, HistorySource historySource,
+                                                        MapperService mapperService, long startingSeqNo) throws IOException;
 
     /**
      * Gets the minimum retained sequence number for this engine.
@@ -1819,7 +1821,8 @@ public IndexCommit getIndexCommit() {
         }
     }
 
-    public void onSettingsChanged() {
+    public void onSettingsChanged(TimeValue translogRetentionAge, ByteSizeValue translogRetentionSize, long softDeletesRetentionOps) {
+
     }
 
     /**
@@ -1953,4 +1956,11 @@ public interface TranslogRecoveryRunner {
      * to advance this marker to at least the given sequence number.
      */
     public abstract void advanceMaxSeqNoOfUpdatesOrDeletes(long maxSeqNoOfUpdatesOnPrimary);
+
+    /**
+     * Whether we should read history operations from translog or Lucene index
+     */
+    public enum HistorySource {
+        TRANSLOG, INDEX
+    }
 }