Implemented exponential backoff and max retry with resumable uploads

firebase-storage/CHANGELOG.md

@@ -1,3 +1,7 @@
+# 20.0.3
+- [fixed] Fixed an issue that caused infinite number of retries with no exponential
+  backoff for `uploadChunk`
+
 # 19.2.2
 - [fixed] Fixed an issue that caused the SDK to report incorrect values for
   "getTotalByteCount()" after a download was paused and resumed.

firebase-storage/src/main/java/com/google/firebase/storage/FirebaseStorage.java

@@ -53,6 +53,7 @@ public class FirebaseStorage {
   @Nullable private final Provider<InternalAppCheckTokenProvider> mAppCheckProvider;
   @Nullable private final String mBucketName;
   private long sMaxUploadRetry = 10 * DateUtils.MINUTE_IN_MILLIS; //  10 * 60 * 1000
+  private long sMaxChunkUploadRetry = DateUtils.MINUTE_IN_MILLIS; //  60 * 1000
   private long sMaxDownloadRetry = 10 * DateUtils.MINUTE_IN_MILLIS; //  10 * 60 * 1000
   private long sMaxQueryRetry = 2 * DateUtils.MINUTE_IN_MILLIS; //  2 * 60 * 1000
 
@@ -226,6 +227,25 @@ public void setMaxUploadRetryTimeMillis(long maxTransferRetryMillis) {
     sMaxUploadRetry = maxTransferRetryMillis;
   }
 
+  /**
+   * Returns the maximum time to retry sending a chunk if a failure occurs
+   *
+   * @return maximum time in milliseconds. Defaults to 1 minute.
+   */
+  public long getMaxChunkUploadRetry() {
+    return sMaxChunkUploadRetry;
+  }
+
+  /**
+   * Sets the maximum time to retry sending a chunk if a failure occurs
+   *
+   * @param maxChunkRetryMillis the maximum time in milliseconds. Defaults to 1 minute (60,000
+   *     milliseconds).
+   */
+  public void setMaxChunkUploadRetry(long maxChunkRetryMillis) {
+    sMaxChunkUploadRetry = maxChunkRetryMillis;
+  }
+
   /**
    * Returns the maximum time to retry operations other than upload and download if a failure
    * occurs.

firebase-storage/src/main/java/com/google/firebase/storage/StorageTaskScheduler.java

@@ -62,6 +62,10 @@ public class StorageTaskScheduler {
     CALLBACK_QUEUE_EXECUTOR.allowCoreThreadTimeOut(true);
   }
 
+  public static void setCallbackQueueKeepAlive(long keepAliveTime, TimeUnit timeUnit) {
+    CALLBACK_QUEUE_EXECUTOR.setKeepAliveTime(keepAliveTime, timeUnit);
+  }
+
   public static StorageTaskScheduler getInstance() {
     return sInstance;
   }

firebase-storage/src/main/java/com/google/firebase/storage/UploadTask.java

@@ -14,6 +14,8 @@
 
 package com.google.firebase.storage;
 
+import static com.google.firebase.storage.internal.ExponentialBackoffSender.RND_MAX;
+
 import android.content.ContentResolver;
 import android.content.Context;
 import android.net.Uri;
@@ -25,10 +27,14 @@
 import androidx.annotation.VisibleForTesting;
 import com.google.android.gms.common.api.Status;
 import com.google.android.gms.common.internal.Preconditions;
+import com.google.android.gms.common.util.Clock;
+import com.google.android.gms.common.util.DefaultClock;
 import com.google.firebase.appcheck.interop.InternalAppCheckTokenProvider;
 import com.google.firebase.auth.internal.InternalAuthProvider;
 import com.google.firebase.storage.internal.AdaptiveStreamBuffer;
 import com.google.firebase.storage.internal.ExponentialBackoffSender;
+import com.google.firebase.storage.internal.Sleeper;
+import com.google.firebase.storage.internal.SleeperImpl;
 import com.google.firebase.storage.internal.Util;
 import com.google.firebase.storage.network.NetworkRequest;
 import com.google.firebase.storage.network.ResumableUploadByteRequest;
@@ -40,6 +46,7 @@
 import java.io.FileNotFoundException;
 import java.io.IOException;
 import java.io.InputStream;
+import java.util.Random;
 import java.util.concurrent.atomic.AtomicLong;
 import org.json.JSONException;
 
@@ -72,6 +79,13 @@ public class UploadTask extends StorageTask<UploadTask.TaskSnapshot> {
   private volatile Exception mServerException = null;
   private volatile int mResultCode = 0;
   private volatile String mServerStatus;
+  private volatile long maxSleepTime;
+  private static final Random random = new Random();
+  /*package*/ static Sleeper sleeper = new SleeperImpl();
+  /*package*/ static Clock clock = DefaultClock.getInstance();
+  private int sleepTime =
+      0; // TODO(mtewani): Make it so that the send is 0,1,2,4,8,... and start at 0
+  private final int sleepInterval = 1000;
 
   UploadTask(StorageReference targetRef, StorageMetadata metadata, byte[] bytes) {
     Preconditions.checkNotNull(targetRef);
@@ -89,6 +103,7 @@ public class UploadTask extends StorageTask<UploadTask.TaskSnapshot> {
         new AdaptiveStreamBuffer(new ByteArrayInputStream(bytes), PREFERRED_CHUNK_SIZE);
     this.mIsStreamOwned = true;
 
+    this.maxSleepTime = storage.getMaxChunkUploadRetry();
     mSender =
         new ExponentialBackoffSender(
             storage.getApp().getApplicationContext(),
@@ -110,6 +125,7 @@ public class UploadTask extends StorageTask<UploadTask.TaskSnapshot> {
     this.mAppCheckProvider = storage.getAppCheckProvider();
     this.mUri = file;
     InputStream inputStream = null;
+    this.maxSleepTime = storage.getMaxChunkUploadRetry();
     mSender =
         new ExponentialBackoffSender(
             mStorageRef.getApp().getApplicationContext(),
@@ -173,12 +189,13 @@ public class UploadTask extends StorageTask<UploadTask.TaskSnapshot> {
     this.mStreamBuffer = new AdaptiveStreamBuffer(stream, PREFERRED_CHUNK_SIZE);
     this.mIsStreamOwned = false;
     this.mUri = null;
+    this.maxSleepTime = storage.getMaxChunkUploadRetry();
     mSender =
         new ExponentialBackoffSender(
             mStorageRef.getApp().getApplicationContext(),
             mAuthProvider,
             mAppCheckProvider,
-            mStorageRef.getStorage().getMaxUploadRetryTimeMillis());
+            storage.getMaxUploadRetryTimeMillis());
   }
 
   /** @return the target of the upload. */
@@ -321,15 +338,18 @@ private boolean shouldContinue() {
     }
 
     boolean inErrorState = mServerException != null || mResultCode < 200 || mResultCode >= 300;
+    long deadLine = clock.elapsedRealtime() + this.maxSleepTime;
+    long currentTime = clock.elapsedRealtime() + this.sleepTime;
     // we attempt to recover by calling recoverStatus(true)
-    if (inErrorState && !recoverStatus(true)) {
-      // we failed to recover.
-      if (serverStateValid()) {
-        tryChangeState(INTERNAL_STATE_FAILURE, false);
+    if (inErrorState) {
+      if (currentTime > deadLine || !recoverStatus(true)) {
+        if (serverStateValid()) {
+          tryChangeState(INTERNAL_STATE_FAILURE, false);
+        }
+        return false;
       }
-      return false;
+      sleepTime = Math.max(sleepTime * 2, sleepTime + (sleepInterval * 2));
     }
-
     return true;
   }
 
@@ -410,6 +430,24 @@ private boolean recoverStatus(boolean withRetry) {
     return true;
   }
 
+  private boolean delaySend(NetworkRequest request) {
+    try {
+      Log.d(TAG, "Waiting " + sleepTime + " milliseconds");
+      sleeper.sleep(sleepTime + random.nextInt(RND_MAX));
+    } catch (InterruptedException e) {
+      Log.w(TAG, "thread interrupted during exponential backoff.");
+
+      Thread.currentThread().interrupt();
+      mServerException = e;
+      return false;
+    }
+    boolean sendRes = send(request);
+    if (sendRes) {
+      sleepTime = 0;
+    }
+    return sendRes;
+  }
+
   private void uploadChunk() {
     try {
       mStreamBuffer.fill(mCurrentChunkSize);
@@ -425,7 +463,7 @@ private void uploadChunk() {
               bytesToUpload,
               mStreamBuffer.isFinished());
 
-      if (!send(uploadRequest)) {
+      if (!delaySend(uploadRequest)) {
         mCurrentChunkSize = PREFERRED_CHUNK_SIZE;
         Log.d(TAG, "Resetting chunk size to " + mCurrentChunkSize);
         return;

firebase-storage/src/main/java/com/google/firebase/storage/network/NetworkRequest.java

@@ -106,7 +106,7 @@ public static Uri getBaseUrl(@Nullable EmulatedServiceSettings emulatorSettings)
       return Uri.parse(
           "http://" + emulatorSettings.getHost() + ":" + emulatorSettings.getPort() + "/v0");
     } else {
-      return Uri.parse("https://firebasestorage.googleapis.com/v0");
+      return PROD_BASE_URL;
     }
   }
 

firebase-storage/src/test/java/com/google/firebase/storage/TestUtil.java

@@ -32,6 +32,12 @@
 public class TestUtil {
 
   static FirebaseApp createApp() {
+    // Many tests require you to call the callback on the same thread that was initially
+    // instantiated.
+    // With the 5 second keepalive, after 5 seconds, the thread will get killed and eventually a new
+    // one will be created.
+    // Therefore causing many of the tests to fail
+    StorageTaskScheduler.setCallbackQueueKeepAlive(90, TimeUnit.SECONDS);
     return FirebaseApp.initializeApp(
         ApplicationProvider.getApplicationContext(),
         new FirebaseOptions.Builder()
@@ -148,6 +154,6 @@ static void await(Task<?> task, int timeout, TimeUnit timeUnit) throws Interrupt
    * Tasks to be executed.
    */
   static void await(Task<?> task) throws InterruptedException {
-    await(task, 3, TimeUnit.SECONDS);
+    await(task, 10, TimeUnit.SECONDS);
   }
 }

firebase-storage/src/test/java/com/google/firebase/storage/UploadTest.java

@@ -100,6 +100,36 @@ public void smallTextUpload() throws Exception {
     TestUtil.verifyTaskStateChanges("smallTextUpload", task.getResult().toString());
   }
 
+  /**
+   * This test will replicate uploadChunk() returning 500's and test to make sure the retries are
+   * limited and using exponential backoff. If the maxretry limit is not checked, then the await
+   * task will time out.
+   *
+   * @throws Exception
+   */
+  @Test
+  public void fileUploadWith500() throws Exception {
+
+    System.out.println("Starting test fileUploadWith500.");
+
+    MockConnectionFactory factory = NetworkLayerMock.ensureNetworkMock("fileUploadWith500", true);
+
+    String filename = TEST_ASSET_ROOT + "image.jpg";
+    ClassLoader classLoader = UploadTest.class.getClassLoader();
+    InputStream imageStream = classLoader.getResourceAsStream(filename);
+    Uri sourceFile = Uri.parse("file://" + filename);
+
+    ContentResolver resolver = ApplicationProvider.getApplicationContext().getContentResolver();
+    Shadows.shadowOf(resolver).registerInputStream(sourceFile, imageStream);
+
+    Task<StringBuilder> task = TestUploadHelper.fileUpload(sourceFile, "image.jpg");
+
+    TestUtil.await(task, 2, TimeUnit.MINUTES);
+
+    factory.verifyOldMock();
+    TestUtil.verifyTaskStateChanges("fileUploadWith500", task.getResult().toString());
+  }
+
   @Test
   public void cantUploadToRoot() throws Exception {
     System.out.println("Starting test cantUploadToRoot.");
@@ -130,7 +160,7 @@ public void cantUploadToRoot() throws Exception {
         });
 
     // TODO(mrschmidt): Lower the timeout
-    TestUtil.await(task, 300, TimeUnit.SECONDS);
+    TestUtil.await(task, 7, TimeUnit.MINUTES);
 
     try {
       task.getResult();
@@ -466,7 +496,7 @@ public void fileUploadRecovery() throws Exception {
 
     Task<StringBuilder> task = TestUploadHelper.fileUpload(sourceFile, "flubbertest.jpg");
 
-    TestUtil.await(task, 5, TimeUnit.SECONDS);
+    TestUtil.await(task, 30, TimeUnit.SECONDS);
 
     factory.verifyOldMock();
     TestUtil.verifyTaskStateChanges("fileUploadRecovery", task.getResult().toString());
@@ -489,7 +519,7 @@ public void fileUploadNoRecovery() throws Exception {
 
     Task<StringBuilder> task = TestUploadHelper.fileUpload(sourceFile, "flubbertest.jpg");
 
-    TestUtil.await(task, 5, TimeUnit.SECONDS);
+    TestUtil.await(task);
 
     factory.verifyOldMock();
     TestUtil.verifyTaskStateChanges("fileUploadNoRecovery", task.getResult().toString());