@Override
public void messageReceived(RecoveryCleanFilesRequest request, TransportChannel channel)
throws Exception {
try (RecoveriesCollection.StatusRef statusRef =
onGoingRecoveries.getStatusSafe(request.recoveryId(), request.shardId())) {
final RecoveryStatus recoveryStatus = statusRef.status();
recoveryStatus.state().getTranslog().totalOperations(request.totalTranslogOps());
// first, we go and move files that were created with the recovery id suffix to
// the actual names, its ok if we have a corrupted index here, since we have replicas
// to recover from in case of a full cluster shutdown just when this code executes...
recoveryStatus
.indexShard()
.deleteShardState(); // we have to delete it first since even if we fail to rename the
// shard might be invalid
recoveryStatus.renameAllTempFiles();
final Store store = recoveryStatus.store();
// now write checksums
recoveryStatus.legacyChecksums().write(store);
Store.MetadataSnapshot sourceMetaData = request.sourceMetaSnapshot();
try {
store.cleanupAndVerify("recovery CleanFilesRequestHandler", sourceMetaData);
} catch (CorruptIndexException
| IndexFormatTooNewException
| IndexFormatTooOldException ex) {
// this is a fatal exception at this stage.
// this means we transferred files from the remote that have not be checksummed and they
// are
// broken. We have to clean up this shard entirely, remove all files and bubble it up to
// the
// source shard since this index might be broken there as well? The Source can handle this
// and checks
// its content on disk if possible.
try {
try {
store.removeCorruptionMarker();
} finally {
Lucene.cleanLuceneIndex(store.directory()); // clean up and delete all files
}
} catch (Throwable e) {
logger.debug("Failed to clean lucene index", e);
ex.addSuppressed(e);
}
RecoveryFailedException rfe =
new RecoveryFailedException(
recoveryStatus.state(), "failed to clean after recovery", ex);
recoveryStatus.fail(rfe, true);
throw rfe;
} catch (Exception ex) {
RecoveryFailedException rfe =
new RecoveryFailedException(
recoveryStatus.state(), "failed to clean after recovery", ex);
recoveryStatus.fail(rfe, true);
throw rfe;
}
channel.sendResponse(TransportResponse.Empty.INSTANCE);
}
}
