Fix Broken Index Shard Snapshot File Preventing Snapshot Creation

server/src/main/java/org/elasticsearch/index/snapshots/blobstore/BlobStoreIndexShardSnapshots.java

@@ -37,10 +37,10 @@
 import static java.util.Collections.unmodifiableMap;
 
 /**
- * Contains information about all snapshot for the given shard in repository
+ * Contains information about all snapshots for the given shard in repository
  * <p>
  * This class is used to find files that were already snapshotted and clear out files that no longer referenced by any
- * snapshots
+ * snapshots.
  */
 public class BlobStoreIndexShardSnapshots implements Iterable<SnapshotFiles>, ToXContentFragment {
 

server/src/main/java/org/elasticsearch/repositories/blobstore/BlobStoreRepository.java

@@ -123,27 +123,28 @@
  * <pre>
  * {@code
  *   STORE_ROOT
- *   |- index-N           - list of all snapshot ids and the indices belonging to each snapshot, N is the generation of the file
+ *   |- index-N           - JSON serialized {@link RepositoryData} containing a list of all snapshot ids and the indices belonging to
+ *   |                      each snapshot, N is the generation of the file
  *   |- index.latest      - contains the numeric value of the latest generation of the index file (i.e. N from above)
  *   |- incompatible-snapshots - list of all snapshot ids that are no longer compatible with the current version of the cluster
- *   |- snap-20131010 - JSON serialized Snapshot for snapshot "20131010"
- *   |- meta-20131010.dat - JSON serialized MetaData for snapshot "20131010" (includes only global metadata)
- *   |- snap-20131011 - JSON serialized Snapshot for snapshot "20131011"
- *   |- meta-20131011.dat - JSON serialized MetaData for snapshot "20131011"
+ *   |- snap-20131010.dat - SMILE serialized {@link SnapshotInfo} for snapshot "20131010"
+ *   |- meta-20131010.dat - SMILE serialized {@link MetaData} for snapshot "20131010" (includes only global metadata)
+ *   |- snap-20131011.dat - SMILE serialized {@link SnapshotInfo} for snapshot "20131011"
+ *   |- meta-20131011.dat - SMILE serialized {@link MetaData} for snapshot "20131011"
  *   .....
  *   |- indices/ - data for all indices
  *      |- Ac1342-B_x/ - data for index "foo" which was assigned the unique id of Ac1342-B_x in the repository
- *      |  |- meta-20131010.dat - JSON Serialized IndexMetaData for index "foo"
+ *      |  |- meta-20131010.dat - JSON Serialized {@link IndexMetaData} for index "foo"
  *      |  |- 0/ - data for shard "0" of index "foo"
- *      |  |  |- __1 \
- *      |  |  |- __2 |
- *      |  |  |- __3 |- files from different segments see snapshot-* for their mappings to real segment files
- *      |  |  |- __4 |
- *      |  |  |- __5 /
+ *      |  |  |- __1                      \  (files with numeric names were created by older ES versions)
+ *      |  |  |- __2                      |
+ *      |  |  |- __VPO5oDMVT5y4Akv8T_AO_A |- files from different segments see snap-* for their mappings to real segment files
+ *      |  |  |- __1gbJy18wS_2kv1qI7FgKuQ |
+ *      |  |  |- __R8JvZAHlSMyMXyZc2SS8Zg /
  *      |  |  .....
- *      |  |  |- snap-20131010.dat - JSON serialized BlobStoreIndexShardSnapshot for snapshot "20131010"
- *      |  |  |- snap-20131011.dat - JSON serialized BlobStoreIndexShardSnapshot for snapshot "20131011"
- *      |  |  |- list-123 - JSON serialized BlobStoreIndexShardSnapshot for snapshot "20131011"
+ *      |  |  |- snap-20131010.dat - SMILE serialized {@link BlobStoreIndexShardSnapshot} for snapshot "20131010"
+ *      |  |  |- snap-20131011.dat - SMILE serialized {@link BlobStoreIndexShardSnapshot} for snapshot "20131011"
+ *      |  |  |- index-123 - SMILE serialized {@link BlobStoreIndexShardSnapshots} for the shard
  *      |  |
  *      |  |- 1/ - data for shard "1" of index "foo"
  *      |  |  |- __1
@@ -1142,7 +1143,7 @@ protected Tuple<BlobStoreIndexShardSnapshots, Integer> buildBlobStoreIndexShardS
                     logger.warn(() -> new ParameterizedMessage("failed to read commit point [{}]", name), e);
                 }
             }
-            return new Tuple<>(new BlobStoreIndexShardSnapshots(snapshots), -1);
+            return new Tuple<>(new BlobStoreIndexShardSnapshots(snapshots), latest);
         }
     }
 

server/src/test/java/org/elasticsearch/snapshots/SharedClusterSnapshotRestoreIT.java

@@ -2963,6 +2963,108 @@ public void testRestoreSnapshotWithCorruptedIndexMetadata() throws Exception {
         assertAcked(client().admin().cluster().prepareDeleteSnapshot("test-repo", snapshotInfo.snapshotId().getName()).get());
     }
 
+    /**
+     * Tests that a shard snapshot with a corrupted shard index file can still be used for restore and incremental snapshots.
+     */
+    public void testSnapshotWithCorruptedShardIndexFile() throws Exception {
+        final Client client = client();
+        final Path repo = randomRepoPath();
+        final String indexName = "test-idx";
+        final int nDocs = randomIntBetween(1, 10);
+
+        logger.info("-->  creating index [{}] with [{}] documents in it", indexName, nDocs);
+        assertAcked(prepareCreate(indexName).setSettings(Settings.builder()
+            .put(SETTING_NUMBER_OF_SHARDS, 1).put(SETTING_NUMBER_OF_REPLICAS, 0)));
+
+        final IndexRequestBuilder[] documents = new IndexRequestBuilder[nDocs];
+        for (int j = 0; j < nDocs; j++) {
+            documents[j] = client.prepareIndex(indexName, "_doc").setSource("foo", "bar");
+        }
+        indexRandom(true, documents);
+        flushAndRefresh();
+
+        logger.info("-->  creating repository");
+        assertAcked(client().admin().cluster().preparePutRepository("test-repo")
+            .setType("fs")
+            .setSettings(Settings.builder()
+                .put("location", repo)));
+
+        final String snapshot1 = "test-snap-1";
+        logger.info("-->  creating snapshot [{}]", snapshot1);
+        final SnapshotInfo snapshotInfo = client().admin().cluster().prepareCreateSnapshot("test-repo", snapshot1)
+            .setWaitForCompletion(true)
+            .get()
+            .getSnapshotInfo();
+        assertThat(snapshotInfo.failedShards(), equalTo(0));
+        assertThat(snapshotInfo.successfulShards(), equalTo(snapshotInfo.totalShards()));
+        assertThat(snapshotInfo.indices(), hasSize(1));
+
+        RepositoriesService service = internalCluster().getInstance(RepositoriesService.class, internalCluster().getMasterName());
+        Repository repository = service.repository("test-repo");
+
+        final RepositoryData repositoryData = getRepositoryData(repository);
+        final Map<String, IndexId> indexIds = repositoryData.getIndices();
+        assertThat(indexIds.size(), equalTo(1));
+
+        final IndexId corruptedIndex = indexIds.get(indexName);
+        final Path shardIndexFile = repo.resolve("indices")
+            .resolve(corruptedIndex.getId()).resolve("0")
+            .resolve("index-0");
+
+        logger.info("-->  truncating shard index file [{}]", shardIndexFile);
+        try (SeekableByteChannel outChan = Files.newByteChannel(shardIndexFile, StandardOpenOption.WRITE)) {
+            outChan.truncate(randomInt(10));
+        }
+
+        logger.info("-->  verifying snapshot state for [{}]", snapshot1);
+        List<SnapshotInfo> snapshotInfos = client().admin().cluster().prepareGetSnapshots("test-repo").get().getSnapshots();
+        assertThat(snapshotInfos.size(), equalTo(1));
+        assertThat(snapshotInfos.get(0).state(), equalTo(SnapshotState.SUCCESS));
+        assertThat(snapshotInfos.get(0).snapshotId().getName(), equalTo(snapshot1));
+
+        logger.info("-->  deleting index [{}]", indexName);
+        assertAcked(client().admin().indices().prepareDelete(indexName));
+
+        logger.info("-->  restoring snapshot [{}]", snapshot1);
+        client().admin().cluster().prepareRestoreSnapshot("test-repo", snapshot1)
+            .setRestoreGlobalState(randomBoolean())
+            .setWaitForCompletion(true)
+            .get();
+        ensureGreen();
+
+        assertHitCount(client().prepareSearch(indexName).setSize(0).get(), nDocs);
+
+        logger.info("-->  indexing [{}] more documents into [{}]", nDocs, indexName);
+        for (int j = 0; j < nDocs; j++) {
+            documents[j] = client.prepareIndex(indexName, "_doc").setSource("foo2", "bar2");
+        }
+        indexRandom(true, documents);
+
+        final String snapshot2 = "test-snap-2";
+        logger.info("-->  creating snapshot [{}]", snapshot2);
+        final SnapshotInfo snapshotInfo2 = client().admin().cluster().prepareCreateSnapshot("test-repo", snapshot2)
+            .setWaitForCompletion(true)
+            .get()
+            .getSnapshotInfo();
+        assertThat(snapshotInfo2.state(), equalTo(SnapshotState.SUCCESS));
+        assertThat(snapshotInfo2.failedShards(), equalTo(0));
+        assertThat(snapshotInfo2.successfulShards(), equalTo(snapshotInfo.totalShards()));
+        assertThat(snapshotInfo2.indices(), hasSize(1));
+
+        logger.info("-->  deleting index [{}]", indexName);
+        assertAcked(client().admin().indices().prepareDelete(indexName));
+
+        logger.info("-->  restoring snapshot [{}]", snapshot2);
+        client().admin().cluster().prepareRestoreSnapshot("test-repo", snapshot2)
+            .setRestoreGlobalState(randomBoolean())
+            .setWaitForCompletion(true)
+            .get();
+
+        ensureGreen();
+
+        assertHitCount(client().prepareSearch(indexName).setSize(0).get(), 2 * nDocs);
+    }
+
     public void testCannotCreateSnapshotsWithSameName() throws Exception {
         final String repositoryName = "test-repo";
         final String snapshotName = "test-snap";