use of com.github.ambry.protocol.AdminRequest in project ambry by linkedin.
the class AmbryRequestsTest method doCatchupStatusTest.
// catchupStatusSuccessTest() and catchupStatusFailureTest() helpers
/**
* Does the test for {@link AdminRequestOrResponseType#CatchupStatus} by checking that the request is correctly
* deserialized in {@link AmbryRequests} and the necessary info obtained from {@link ReplicationManager}.
* @param id the {@link PartitionId} to disable replication on. Can be {@code null}.
* @param acceptableLagInBytes the value of acceptable lag to set in the {@link CatchupStatusAdminRequest}.
* @param numReplicasCaughtUpPerPartition the value of num replicas caught up per partition to set in the
* {@link CatchupStatusAdminRequest}.
* @param expectedServerErrorCode the {@link ServerErrorCode} expected in the response.
* @param expectedIsCaughtUp the expected return from {@link CatchupStatusAdminResponse#isCaughtUp()}.
* @throws InterruptedException
* @throws IOException
*/
private void doCatchupStatusTest(PartitionId id, long acceptableLagInBytes, short numReplicasCaughtUpPerPartition, ServerErrorCode expectedServerErrorCode, boolean expectedIsCaughtUp) throws InterruptedException, IOException {
int correlationId = TestUtils.RANDOM.nextInt();
String clientId = UtilsTest.getRandomString(10);
AdminRequest adminRequest = new AdminRequest(AdminRequestOrResponseType.CatchupStatus, id, correlationId, clientId);
CatchupStatusAdminRequest catchupStatusRequest = new CatchupStatusAdminRequest(acceptableLagInBytes, numReplicasCaughtUpPerPartition, adminRequest);
Response response = sendRequestGetResponse(catchupStatusRequest, expectedServerErrorCode);
assertTrue("Response not of type CatchupStatusAdminResponse", response instanceof CatchupStatusAdminResponse);
CatchupStatusAdminResponse adminResponse = (CatchupStatusAdminResponse) response;
assertEquals("Catchup status not as expected", expectedIsCaughtUp, adminResponse.isCaughtUp());
}
use of com.github.ambry.protocol.AdminRequest in project ambry by linkedin.
the class AmbryRequestsTest method stopBlobStoreSuccessTest.
/**
* Tests for the response received on a {@link BlobStoreControlAdminRequest} for successful case
* @throws InterruptedException
* @throws IOException
*/
@Test
public void stopBlobStoreSuccessTest() throws InterruptedException, IOException {
List<? extends PartitionId> partitionIds = clusterMap.getAllPartitionIds();
PartitionId id = partitionIds.get(0);
List<? extends ReplicaId> replicaIds = id.getReplicaIds();
assertTrue("This test needs more than one replica for the first partition to work", replicaIds.size() > 1);
int correlationId = TestUtils.RANDOM.nextInt();
String clientId = UtilsTest.getRandomString(10);
long acceptableLagInBytes = 0;
short numReplicasCaughtUpPerPartition = 3;
replicationManager.reset();
replicationManager.controlReplicationReturnVal = true;
generateLagOverrides(0, acceptableLagInBytes);
// stop BlobStore
AdminRequest adminRequest = new AdminRequest(AdminRequestOrResponseType.BlobStoreControl, id, correlationId, clientId);
BlobStoreControlAdminRequest stopBlobStoreAdminRequest = new BlobStoreControlAdminRequest(numReplicasCaughtUpPerPartition, false, adminRequest);
Response response = sendRequestGetResponse(stopBlobStoreAdminRequest, ServerErrorCode.No_Error);
assertTrue("Response not of type AdminResponse", response instanceof AdminResponse);
// test APIs are called in the process of stopping BlobStore
assertEquals("Partition disabled for compaction not as expected", id, storageManager.compactionDisabledPartitionId);
assertEquals("Origins list should be empty", true, replicationManager.originsVal.isEmpty());
assertEquals("Replication on given BlobStore should be disabled", false, replicationManager.enableVal);
assertEquals("Partition shutdown not as expected", id, storageManager.shutdownPartitionId);
}
use of com.github.ambry.protocol.AdminRequest in project ambry by linkedin.
the class AmbryRequestsTest method sendAndVerifyRequestControlRequest.
/**
* Sends and verifies that a {@link AdminRequestOrResponseType#RequestControl} request received the error code
* expected.
* @param toControl the {@link AdminRequestOrResponseType#RequestControl} to control.
* @param enable {@code true} if {@code toControl} needs to be enabled. {@code false} otherwise.
* @param id the {@link PartitionId} to send the request for. Can be {@code null}.
* @param expectedServerErrorCode the {@link ServerErrorCode} expected in the response.
* @throws InterruptedException
* @throws IOException
*/
private void sendAndVerifyRequestControlRequest(RequestOrResponseType toControl, boolean enable, PartitionId id, ServerErrorCode expectedServerErrorCode) throws InterruptedException, IOException {
int correlationId = TestUtils.RANDOM.nextInt();
String clientId = UtilsTest.getRandomString(10);
AdminRequest adminRequest = new AdminRequest(AdminRequestOrResponseType.RequestControl, id, correlationId, clientId);
RequestControlAdminRequest controlRequest = new RequestControlAdminRequest(toControl, enable, adminRequest);
Response response = sendRequestGetResponse(controlRequest, expectedServerErrorCode);
assertTrue("Response not of type AdminResponse", response instanceof AdminResponse);
}
use of com.github.ambry.protocol.AdminRequest in project ambry by linkedin.
the class ServerTestUtil method endToEndReplicationWithMultiNodeMultiPartitionMultiDCTest.
static void endToEndReplicationWithMultiNodeMultiPartitionMultiDCTest(String sourceDatacenter, String sslEnabledDatacenters, PortType portType, MockCluster cluster, MockNotificationSystem notificationSystem, Properties routerProps) throws Exception {
Properties props = new Properties();
props.setProperty("router.hostname", "localhost");
props.setProperty("router.datacenter.name", sourceDatacenter);
props.setProperty("router.put.request.parallelism", "1");
props.setProperty("router.put.success.target", "1");
props.setProperty("clustermap.cluster.name", "test");
props.setProperty("clustermap.datacenter.name", sourceDatacenter);
props.setProperty("clustermap.host.name", "localhost");
props.setProperty("kms.default.container.key", TestUtils.getRandomKey(32));
props.putAll(routerProps);
VerifiableProperties verifiableProperties = new VerifiableProperties(props);
AccountService accountService = new InMemAccountService(false, true);
Router router = new NonBlockingRouterFactory(verifiableProperties, cluster.getClusterMap(), notificationSystem, getSSLFactoryIfRequired(verifiableProperties), accountService).getRouter();
int numberOfRequestsToSend = 15;
int numberOfVerifierThreads = 3;
final LinkedBlockingQueue<Payload> payloadQueue = new LinkedBlockingQueue<Payload>();
final AtomicReference<Exception> exceptionRef = new AtomicReference<>(null);
final CountDownLatch callbackLatch = new CountDownLatch(numberOfRequestsToSend);
List<Future<String>> putFutures = new ArrayList<>(numberOfRequestsToSend);
short accountId = Utils.getRandomShort(TestUtils.RANDOM);
short containerId = Utils.getRandomShort(TestUtils.RANDOM);
for (int i = 0; i < numberOfRequestsToSend; i++) {
int size = new Random().nextInt(5000);
final BlobProperties properties = new BlobProperties(size, "service1", "owner id check", "image/jpeg", false, TestUtils.TTL_SECS, cluster.time.milliseconds(), accountId, containerId, false, null, null, null);
final byte[] metadata = new byte[new Random().nextInt(1000)];
final byte[] blob = new byte[size];
TestUtils.RANDOM.nextBytes(metadata);
TestUtils.RANDOM.nextBytes(blob);
Future<String> future = router.putBlob(properties, metadata, new ByteBufferReadableStreamChannel(ByteBuffer.wrap(blob)), new PutBlobOptionsBuilder().build(), new Callback<String>() {
@Override
public void onCompletion(String result, Exception exception) {
if (exception == null) {
payloadQueue.add(new Payload(properties, metadata, blob, result));
} else {
exceptionRef.set(exception);
}
callbackLatch.countDown();
}
}, QUOTA_CHARGE_EVENT_LISTENER);
putFutures.add(future);
}
for (Future<String> future : putFutures) {
future.get(20, TimeUnit.SECONDS);
}
assertTrue("Did not receive all callbacks in time", callbackLatch.await(1, TimeUnit.SECONDS));
if (exceptionRef.get() != null) {
throw exceptionRef.get();
}
// put away for future use
Payload payload1 = payloadQueue.peek();
MockClusterMap clusterMap = cluster.getClusterMap();
BlobId blobId1 = new BlobId(payload1.blobId, clusterMap);
assertEquals("Did not put expected number of blobs", numberOfRequestsToSend, payloadQueue.size());
Properties sslProps = new Properties();
sslProps.putAll(routerProps);
sslProps.setProperty("clustermap.ssl.enabled.datacenters", sslEnabledDatacenters);
sslProps.setProperty("clustermap.cluster.name", "test");
sslProps.setProperty("clustermap.datacenter.name", sourceDatacenter);
sslProps.setProperty("clustermap.host.name", "localhost");
sslProps.setProperty("connectionpool.read.timeout.ms", "15000");
VerifiableProperties vProps = new VerifiableProperties(sslProps);
ConnectionPool connectionPool = new BlockingChannelConnectionPool(new ConnectionPoolConfig(vProps), new SSLConfig(vProps), new ClusterMapConfig(vProps), new MetricRegistry());
CountDownLatch verifierLatch = new CountDownLatch(numberOfVerifierThreads);
AtomicInteger totalRequests = new AtomicInteger(numberOfRequestsToSend);
AtomicInteger verifiedRequests = new AtomicInteger(0);
AtomicBoolean cancelTest = new AtomicBoolean(false);
for (int i = 0; i < numberOfVerifierThreads; i++) {
Thread thread = new Thread(new Verifier(payloadQueue, verifierLatch, totalRequests, verifiedRequests, cluster.getClusterMap(), cancelTest, portType, connectionPool, notificationSystem, cluster.time));
thread.start();
}
assertTrue("Did not verify in 2 minutes", verifierLatch.await(2, TimeUnit.MINUTES));
assertEquals(totalRequests.get(), verifiedRequests.get());
BlobIdFactory blobIdFactory = new BlobIdFactory(clusterMap);
MockDataNodeId dataNodeId = clusterMap.getDataNodes().get(0);
Port port = new Port(portType == PortType.PLAINTEXT ? dataNodeId.getPort() : dataNodeId.getSSLPort(), portType);
ConnectedChannel channel = connectionPool.checkOutConnection("localhost", port, 10000);
PartitionId partitionId = blobId1.getPartition();
// stop the store via AdminRequest
System.out.println("Begin to stop a BlobStore");
AdminRequest adminRequest = new AdminRequest(AdminRequestOrResponseType.BlobStoreControl, partitionId, 1, "clientid2");
BlobStoreControlAdminRequest controlRequest = new BlobStoreControlAdminRequest((short) 0, BlobStoreControlAction.StopStore, adminRequest);
DataInputStream stream = channel.sendAndReceive(controlRequest).getInputStream();
AdminResponse adminResponse = AdminResponse.readFrom(stream);
releaseNettyBufUnderneathStream(stream);
assertEquals("Stop store admin request should succeed", ServerErrorCode.No_Error, adminResponse.getError());
// put a blob on a stopped store, which should fail
byte[] userMetadata = new byte[1000];
byte[] data = new byte[3187];
BlobProperties properties = new BlobProperties(3187, "serviceid1", accountId, containerId, false, cluster.time.milliseconds());
BlobId blobId2 = new BlobId(CommonTestUtils.getCurrentBlobIdVersion(), BlobId.BlobIdType.NATIVE, clusterMap.getLocalDatacenterId(), accountId, containerId, partitionId, false, BlobId.BlobDataType.DATACHUNK);
PutRequest putRequest2 = new PutRequest(1, "clientId2", blobId2, properties, ByteBuffer.wrap(userMetadata), Unpooled.wrappedBuffer(data), properties.getBlobSize(), BlobType.DataBlob, null);
DataInputStream putResponseStream = channel.sendAndReceive(putRequest2).getInputStream();
PutResponse response2 = PutResponse.readFrom(putResponseStream);
releaseNettyBufUnderneathStream(putResponseStream);
assertEquals("Put blob on stopped store should fail", ServerErrorCode.Replica_Unavailable, response2.getError());
// get a blob properties on a stopped store, which should fail
ArrayList<BlobId> ids = new ArrayList<>();
ids.add(blobId1);
ArrayList<PartitionRequestInfo> partitionRequestInfoList = new ArrayList<>();
PartitionRequestInfo partitionRequestInfo = new PartitionRequestInfo(partitionId, ids);
partitionRequestInfoList.add(partitionRequestInfo);
GetRequest getRequest1 = new GetRequest(1, "clientId1", MessageFormatFlags.BlobProperties, partitionRequestInfoList, GetOption.None);
stream = channel.sendAndReceive(getRequest1).getInputStream();
GetResponse resp1 = GetResponse.readFrom(stream, clusterMap);
assertEquals("Get blob properties on stopped store should fail", ServerErrorCode.Replica_Unavailable, resp1.getPartitionResponseInfoList().get(0).getErrorCode());
releaseNettyBufUnderneathStream(stream);
// delete a blob on a stopped store, which should fail
DeleteRequest deleteRequest = new DeleteRequest(1, "clientId1", blobId1, System.currentTimeMillis());
stream = channel.sendAndReceive(deleteRequest).getInputStream();
DeleteResponse deleteResponse = DeleteResponse.readFrom(stream);
releaseNettyBufUnderneathStream(stream);
assertEquals("Delete blob on stopped store should fail", ServerErrorCode.Replica_Unavailable, deleteResponse.getError());
// start the store via AdminRequest
System.out.println("Begin to restart the BlobStore");
adminRequest = new AdminRequest(AdminRequestOrResponseType.BlobStoreControl, partitionId, 1, "clientId");
controlRequest = new BlobStoreControlAdminRequest((short) 0, BlobStoreControlAction.StartStore, adminRequest);
stream = channel.sendAndReceive(controlRequest).getInputStream();
adminResponse = AdminResponse.readFrom(stream);
releaseNettyBufUnderneathStream(stream);
assertEquals("Start store admin request should succeed", ServerErrorCode.No_Error, adminResponse.getError());
List<? extends ReplicaId> replicaIds = partitionId.getReplicaIds();
for (ReplicaId replicaId : replicaIds) {
// forcibly mark replicas and disks as up.
MockReplicaId mockReplicaId = (MockReplicaId) replicaId;
mockReplicaId.markReplicaDownStatus(false);
((MockDiskId) mockReplicaId.getDiskId()).setDiskState(HardwareState.AVAILABLE, false);
}
// put a blob on a restarted store , which should succeed
putRequest2 = new PutRequest(1, "clientId2", blobId2, properties, ByteBuffer.wrap(userMetadata), Unpooled.wrappedBuffer(data), properties.getBlobSize(), BlobType.DataBlob, null);
putResponseStream = channel.sendAndReceive(putRequest2).getInputStream();
response2 = PutResponse.readFrom(putResponseStream);
releaseNettyBufUnderneathStream(putResponseStream);
assertEquals("Put blob on restarted store should succeed", ServerErrorCode.No_Error, response2.getError());
// verify the put blob has been replicated successfully.
notificationSystem.awaitBlobCreations(blobId2.getID());
// get a blob on a restarted store , which should succeed
ids = new ArrayList<BlobId>();
ids.add(blobId2);
partitionRequestInfoList = new ArrayList<PartitionRequestInfo>();
partitionRequestInfo = new PartitionRequestInfo(partitionId, ids);
partitionRequestInfoList.add(partitionRequestInfo);
GetRequest getRequest2 = new GetRequest(1, "clientId2", MessageFormatFlags.All, partitionRequestInfoList, GetOption.None);
stream = channel.sendAndReceive(getRequest2).getInputStream();
GetResponse resp2 = GetResponse.readFrom(stream, clusterMap);
InputStream responseStream = resp2.getInputStream();
BlobAll blobAll = MessageFormatRecord.deserializeBlobAll(responseStream, blobIdFactory);
byte[] actualBlobData = getBlobDataAndRelease(blobAll.getBlobData());
assertArrayEquals("Content mismatch.", data, actualBlobData);
releaseNettyBufUnderneathStream(stream);
// delete a blob on a restarted store , which should succeed
deleteRequest = new DeleteRequest(1, "clientId2", blobId2, System.currentTimeMillis());
stream = channel.sendAndReceive(deleteRequest).getInputStream();
deleteResponse = DeleteResponse.readFrom(stream);
releaseNettyBufUnderneathStream(stream);
assertEquals("Delete blob on restarted store should succeed", ServerErrorCode.No_Error, deleteResponse.getError());
router.close();
connectionPool.shutdown();
}
use of com.github.ambry.protocol.AdminRequest in project ambry by linkedin.
the class AmbryServerRequests method handleAdminRequest.
/**
* Handles an administration request. These requests can query for or change the internal state of the server.
* @param request the request that needs to be handled.
* @throws InterruptedException if response sending is interrupted.
* @throws IOException if there are I/O errors carrying our the required operation.
*/
@Override
public void handleAdminRequest(NetworkRequest request) throws InterruptedException, IOException {
long requestQueueTime = SystemTime.getInstance().milliseconds() - request.getStartTimeInMs();
long totalTimeSpent = requestQueueTime;
long startTime = SystemTime.getInstance().milliseconds();
DataInputStream requestStream = new DataInputStream(request.getInputStream());
AdminRequest adminRequest = AdminRequest.readFrom(requestStream, clusterMap);
Histogram processingTimeHistogram = null;
Histogram responseQueueTimeHistogram = null;
Histogram responseSendTimeHistogram = null;
Histogram requestTotalTimeHistogram = null;
AdminResponse response = null;
try {
switch(adminRequest.getType()) {
case TriggerCompaction:
metrics.triggerCompactionRequestQueueTimeInMs.update(requestQueueTime);
metrics.triggerCompactionRequestRate.mark();
processingTimeHistogram = metrics.triggerCompactionResponseQueueTimeInMs;
responseQueueTimeHistogram = metrics.triggerCompactionResponseQueueTimeInMs;
responseSendTimeHistogram = metrics.triggerCompactionResponseSendTimeInMs;
requestTotalTimeHistogram = metrics.triggerCompactionRequestTotalTimeInMs;
response = handleTriggerCompactionRequest(adminRequest);
break;
case RequestControl:
metrics.requestControlRequestQueueTimeInMs.update(requestQueueTime);
metrics.requestControlRequestRate.mark();
processingTimeHistogram = metrics.requestControlResponseQueueTimeInMs;
responseQueueTimeHistogram = metrics.requestControlResponseQueueTimeInMs;
responseSendTimeHistogram = metrics.requestControlResponseSendTimeInMs;
requestTotalTimeHistogram = metrics.requestControlRequestTotalTimeInMs;
response = handleRequestControlRequest(requestStream, adminRequest);
break;
case ReplicationControl:
metrics.replicationControlRequestQueueTimeInMs.update(requestQueueTime);
metrics.replicationControlRequestRate.mark();
processingTimeHistogram = metrics.replicationControlResponseQueueTimeInMs;
responseQueueTimeHistogram = metrics.replicationControlResponseQueueTimeInMs;
responseSendTimeHistogram = metrics.replicationControlResponseSendTimeInMs;
requestTotalTimeHistogram = metrics.replicationControlRequestTotalTimeInMs;
response = handleReplicationControlRequest(requestStream, adminRequest);
break;
case CatchupStatus:
metrics.catchupStatusRequestQueueTimeInMs.update(requestQueueTime);
metrics.catchupStatusRequestRate.mark();
processingTimeHistogram = metrics.catchupStatusResponseQueueTimeInMs;
responseQueueTimeHistogram = metrics.catchupStatusResponseQueueTimeInMs;
responseSendTimeHistogram = metrics.catchupStatusResponseSendTimeInMs;
requestTotalTimeHistogram = metrics.catchupStatusRequestTotalTimeInMs;
response = handleCatchupStatusRequest(requestStream, adminRequest);
break;
case BlobStoreControl:
metrics.blobStoreControlRequestQueueTimeInMs.update(requestQueueTime);
metrics.blobStoreControlRequestRate.mark();
processingTimeHistogram = metrics.blobStoreControlRequestQueueTimeInMs;
responseQueueTimeHistogram = metrics.blobStoreControlRequestQueueTimeInMs;
responseSendTimeHistogram = metrics.blobStoreControlResponseSendTimeInMs;
requestTotalTimeHistogram = metrics.blobStoreControlRequestTotalTimeInMs;
response = handleBlobStoreControlRequest(requestStream, adminRequest);
break;
}
} catch (Exception e) {
logger.error("Unknown exception for admin request {}", adminRequest, e);
metrics.unExpectedAdminOperationError.inc();
response = new AdminResponse(adminRequest.getCorrelationId(), adminRequest.getClientId(), ServerErrorCode.Unknown_Error);
switch(adminRequest.getType()) {
case CatchupStatus:
response = new CatchupStatusAdminResponse(false, response);
break;
}
} finally {
long processingTime = SystemTime.getInstance().milliseconds() - startTime;
totalTimeSpent += processingTime;
publicAccessLogger.info("{} {} processingTime {}", adminRequest, response, processingTime);
processingTimeHistogram.update(processingTime);
}
requestResponseChannel.sendResponse(response, request, new ServerNetworkResponseMetrics(responseQueueTimeHistogram, responseSendTimeHistogram, requestTotalTimeHistogram, null, null, totalTimeSpent));
}
Aggregations