Merge dynamic manifest updates to Github (#7)

This change introduces dynamic manifest updates to asset streaming.

Asset streaming describes the directory to be streamed in a manifest, which is a proto definition of all content metadata. This information is sufficient to answer `stat` and `readdir` calls in the FUSE layer without additional round-trips to the workstation.

When a directory is streamed for the first time, the corresponding manifest is created in two steps:
1. The directory is traversed recursively and the inode information of all contained files and directories is written to the manifest.
2. The content of all identified files is processed to generate each file's chunk list. This list is part of the definition of a file in the manifest.
  * The chunk boundaries are identified using our implementation of the FastCDC algorithm.
  * The hash of each chunk is calculated using the BLAKE3 hash function.
  * The length and hash of each chunk is appended to the file's chunk list.

Prior to this change, when the user mounted a workstation directory on a client, the asset streaming server pushed an intermediate manifest to the gamelet as soon as step 1 was completed. At this point, the FUSE client started serving the virtual file system and was ready to answer `stat` and `readdir` calls. In case the FUSE client received any call that required file contents, such as `read`, it would block the caller until the server completed step 2 above and pushed the final manifest to the client. This works well for large directories (> 100GB) with a reasonable number of files (< 100k). But when dealing with millions of tiny files, creating the full manifest can take several minutes.

With this change, we introduce dynamic manifest updates. When the FUSE layer receives an `open` or `readdir` request for a file or directory that is incomplete, it sends an RPC to the workstation about what information is missing from the manifest. The workstation identifies the corresponding file chunker or directory scanner tasks and moves them to the front of the queue. As soon as the task is completed, the workstation pushes an updated intermediate manifest to the client which now includes the information to serve the FUSE request. The queued FUSE request is resumed and returns the result to the caller.

While this does not reduce the required time to build the final manifest, it splits up the work into smaller tasks. This allows us to interrupt the current work and prioritize those tasks which are required to handle an incoming request from the client. While this still takes a round-trip to the workstation plus the processing time for the task, an updated manifest is received within a few seconds, which is much better than blocking for several minutes. 

This latency is only visible when serving data while the manifest is still being created. The situation improves as the manifest creation on the workstation progresses. As soon as the final manifest is pushed, all metadata can be served directly without having to wait for pending tasks.

cdc_fuse_fs/cdc_fuse_fs.cc

@@ -125,12 +125,28 @@ struct Inode {
 // Asset proto -> inode map.
 using InodeMap = std::unordered_map<const AssetProto*, std::shared_ptr<Inode>>;
 
-// Queued request to open a file that has not been processed yet and should be
-// processed once the manifest is updated.
-struct OpenRequest {
-  fuse_req_t req;
-  fuse_ino_t ino;
-  struct fuse_file_info* fi;
+// Queued request that cannot be processed yet and should be processed once the
+// manifest is updated.
+struct QueuedRequest {
+  // The request type that was blocked.
+  enum class Type { kOpen, kOpenDir, kLookup };
+  Type type;
+  std::string rel_path;
+
+  union {
+    // Only valid for type == kOpen or type == kOpenDir.
+    struct Open {
+      fuse_req_t req;
+      fuse_ino_t ino;
+      struct fuse_file_info fi;
+    } open;
+    // Only valid for type == kLookup.
+    struct Lookup {
+      fuse_req_t req;
+      fuse_ino_t parent_ino;
+      const char* name;
+    } lookup;
+  } u;
 };
 
 // Global context. Fuse is based on loose callbacks, so this holds the fs state.
@@ -170,12 +186,13 @@ struct CdcFuseFsContext {
   static thread_local Buffer buffer;
 
   // Configuration client to get configuration updates from the workstation.
-  std::unique_ptr<ConfigStreamClient> config_stream_client_;
+  std::unique_ptr<ConfigStreamClient> config_stream_client;
 
-  // Queue for requests to open files that have not been processed yet.
-  absl::Mutex queued_open_requests_mutex_;
-  std::vector<OpenRequest> queued_open_requests_
-      ABSL_GUARDED_BY(queued_open_requests_mutex_);
+  // Queue for requests to open files or directories that have not been
+  // processed yet.
+  absl::Mutex queued_requests_mutex;
+  std::vector<QueuedRequest> queued_requests
+      ABSL_GUARDED_BY(queued_requests_mutex);
 
   // Identifies whether FUSE consistency should be inspected after manifest
   // update.
@@ -372,6 +389,75 @@ bool ValidateInode(fuse_req_t req, Inode& inode, fuse_ino_t ino) {
   }
   return true;
 }
+
+// Returns the full relative file path for the given |inode|.
+std::string GetRelativePath(const Inode& inode) {
+  if (inode.asset.parent_ino() == FUSE_ROOT_ID)
+    return inode.asset.proto()->name();
+  std::string rel_path = GetRelativePath(GetInode(inode.asset.parent_ino()));
+  absl::StrAppend(&rel_path, "/", inode.asset.proto()->name());
+  return rel_path;
+}
+
+// Asks the server to prioritize the asset at |rel_file_path|.
+void PrioritizeAssetOnServer(const std::string& rel_file_path) {
+  std::vector<std::string> assets{rel_file_path};
+  LOG_INFO("Requesing server to prioritize asset '%s'", rel_file_path);
+  absl::Status status =
+      ctx->config_stream_client->ProcessAssets(std::move(assets));
+  // An error is not critical, but we should log it.
+  if (!status.ok()) {
+    LOG_ERROR(
+        "Failed to request prioritization for asset '%s' from the server: %s",
+        rel_file_path, status.ToString());
+  }
+}
+
+// Queues a CdcFuseOpen request in the list of pending requests. Thread-safe.
+void QueueOpenRequest(const std::string& rel_path, fuse_req_t req,
+                      fuse_ino_t ino, struct fuse_file_info* fi)
+    ABSL_LOCKS_EXCLUDED(ctx->queued_requests_mutex) {
+  QueuedRequest qr{QueuedRequest::Type::kOpen, rel_path};
+  qr.u.open.req = req;
+  qr.u.open.ino = ino;
+  qr.u.open.fi = *fi;
+  {
+    absl::MutexLock lock(&ctx->queued_requests_mutex);
+    ctx->queued_requests.emplace_back(std::move(qr));
+  }
+  PrioritizeAssetOnServer(rel_path);
+}
+
+// Queues a CdcFuseOpenDir request in the list of pending requests. Thread-safe.
+void QueueOpenDirRequest(const std::string& rel_path, fuse_req_t req,
+                         fuse_ino_t ino, struct fuse_file_info* fi)
+    ABSL_LOCKS_EXCLUDED(ctx->queued_requests_mutex) {
+  QueuedRequest qr{QueuedRequest::Type::kOpenDir, rel_path};
+  qr.u.open.req = req;
+  qr.u.open.ino = ino;
+  qr.u.open.fi = *fi;
+  {
+    absl::MutexLock lock(&ctx->queued_requests_mutex);
+    ctx->queued_requests.emplace_back(std::move(qr));
+  }
+  PrioritizeAssetOnServer(rel_path);
+}
+
+// Queues a CdcFuseLookup reuquest in the list of pending requests. Thread-safe.
+void QueueLookupRequest(const std::string& rel_path, fuse_req_t req,
+                        fuse_ino_t parent_ino, const char* name)
+    ABSL_LOCKS_EXCLUDED(ctx->queued_requests_mutex) {
+  QueuedRequest qr{QueuedRequest::Type::kLookup, rel_path};
+  qr.u.lookup.req = req;
+  qr.u.lookup.parent_ino = parent_ino;
+  qr.u.lookup.name = name;
+  {
+    absl::MutexLock lock(&ctx->queued_requests_mutex);
+    ctx->queued_requests.emplace_back(std::move(qr));
+  }
+  PrioritizeAssetOnServer(rel_path);
+}
+
 }  // namespace
 
 // Implementation of the Fuse lookup() method.
@@ -384,6 +470,17 @@ void CdcFuseLookup(fuse_req_t req, fuse_ino_t parent_ino, const char* name)
   if (!ValidateInode(req, parent, parent_ino)) {
     return;
   }
+
+  if (parent.asset.proto()->in_progress()) {
+    // This directory has not been processed yet. Queue up the request and block
+    // until an updated manifest is available.
+    std::string rel_path = GetRelativePath(parent);
+    LOG_INFO("Request to open ino %u queued (file '%s' not ready)", parent_ino,
+             rel_path);
+    QueueLookupRequest(rel_path, req, parent_ino, name);
+    return;
+  }
+
   absl::StatusOr<const AssetProto*> proto = parent.asset.Lookup(name);
   if (!proto.ok()) {
     LOG_ERROR("Lookup of '%s' in ino %u failed: '%s'", name, parent_ino,
@@ -468,13 +565,11 @@ void CdcFuseOpen(fuse_req_t req, fuse_ino_t ino, struct fuse_file_info* fi)
     return;
   }
 
-  if (proto->file_size() > 0 && proto->file_chunks_size() == 0 &&
-      proto->file_indirect_chunks_size() == 0) {
+  if (proto->file_size() > 0 && proto->in_progress()) {
     // This file has not been processed yet. Queue up the request Block until an
     // updated manifest is available.
     LOG_DEBUG("Request to open ino %u queued (file not ready)", ino);
-    absl::MutexLock lock(&ctx->queued_open_requests_mutex_);
-    ctx->queued_open_requests_.push_back({req, ino, fi});
+    QueueOpenRequest(GetRelativePath(inode), req, ino, fi);
     return;
   }
 
@@ -566,6 +661,15 @@ void CdcFuseOpenDir(fuse_req_t req, fuse_ino_t ino, struct fuse_file_info* fi)
     fuse_reply_err(req, ENOTDIR);
     return;
   }
+  if (proto->in_progress()) {
+    // This directory has not been processed yet. Queue up the request until an
+    // updated manifest is available.
+    LOG_DEBUG("Request to open directory '%s' ino %u queued (dir not ready)",
+              proto->name(), ino);
+    QueueOpenDirRequest(GetRelativePath(inode), req, ino, fi);
+    return;
+  }
+
   fuse_reply_open(req, fi);
 }
 
@@ -1462,8 +1566,8 @@ absl::Status SetManifest(const ContentIdProto& manifest_id)
     absl::MutexLock inodes_lock(&ctx->inodes_mutex);
     for (const auto& [proto, inode] : ctx->inodes) {
       // Reset kUpdatedProto to kInitialized. The state was only used for
-      // validation. kUpdated is still needed for clearing kernel caches when a
-      // file is opened.
+      // validation. kUpdated is still needed for clearing kernel caches when
+      // a file is opened.
       assert(inode->IsValid());
       if (inode->IsUpdatedProto() ||
           inode->asset.proto()->type() == AssetProto::DIRECTORY) {
@@ -1474,21 +1578,36 @@ absl::Status SetManifest(const ContentIdProto& manifest_id)
   }
 
   // Process outstanding open requests. Be sure to move the vector because
-  // CdcFuseOpen() might requeue requests.
-  std::vector<OpenRequest> requests;
+  // processing might requeue requests.
+  std::vector<QueuedRequest> requests;
   {
-    absl::MutexLock lock(&ctx->queued_open_requests_mutex_);
-    requests.swap(ctx->queued_open_requests_);
+    absl::MutexLock lock(&ctx->queued_requests_mutex);
+    requests.swap(ctx->queued_requests);
   }
-  for (const OpenRequest request : requests) {
-    LOG_DEBUG("Resuming request to open ino %u", request.ino);
-    CdcFuseOpen(request.req, request.ino, request.fi);
+  for (QueuedRequest& qr : requests) {
+    switch (qr.type) {
+      case QueuedRequest::Type::kLookup:
+        LOG_DEBUG("Resuming request to look up '%s' in '%s' (ino %u)",
+                  qr.u.lookup.name, qr.rel_path, qr.u.lookup.parent_ino);
+        CdcFuseLookup(qr.u.lookup.req, qr.u.lookup.parent_ino,
+                      qr.u.lookup.name);
+        break;
+      case QueuedRequest::Type::kOpen:
+        LOG_DEBUG("Resuming request to open file '%s' (ino %u)", qr.rel_path,
+                  qr.u.open.ino);
+        CdcFuseOpen(qr.u.open.req, qr.u.open.ino, &qr.u.open.fi);
+        break;
+      case QueuedRequest::Type::kOpenDir:
+        LOG_DEBUG("Resuming request to open dir '%s' (ino %u)", qr.rel_path,
+                  qr.u.open.ino);
+        CdcFuseOpenDir(qr.u.open.req, qr.u.open.ino, &qr.u.open.fi);
+        break;
+    }
   }
 
 #ifndef USE_MOCK_LIBFUSE
   // Acknowledge that the manifest id was received and FUSE was updated.
-  absl::Status status =
-      ctx->config_stream_client_->SendManifestAck(manifest_id);
+  absl::Status status = ctx->config_stream_client->SendManifestAck(manifest_id);
   if (!status.ok()) {
     LOG_ERROR("Failed to send ack for manifest '%s'",
               ContentId::ToHexString(manifest_id));
@@ -1500,16 +1619,14 @@ absl::Status SetManifest(const ContentIdProto& manifest_id)
   return absl::OkStatus();
 }
 
-absl::Status StartConfigClient(std::string instance,
-                               std::shared_ptr<grpc::Channel> channel) {
+void SetConfigClient(
+    std::unique_ptr<cdc_ft::ConfigStreamClient> config_client) {
   LOG_DEBUG("Starting configuration client");
   assert(ctx && ctx->initialized);
-  if (ctx->config_stream_client_) {
-    ctx->config_stream_client_.reset();
+  if (ctx->config_stream_client) {
+    ctx->config_stream_client.reset();
   }
-  ctx->config_stream_client_ = std::make_unique<ConfigStreamClient>(
-      std::move(instance), std::move(channel));
-  return absl::OkStatus();
+  ctx->config_stream_client = std::move(config_client);
 }
 
 // Initializes FUSE with a manifest for an empty directory:
@@ -1542,7 +1659,7 @@ absl::Status Run(DataStoreReader* data_store_reader, bool consistency_check) {
   if (res == -1) return MakeStatus("Session loop failed");
   LOG_INFO("Session loop finished.");
 
-  ctx->config_stream_client_->Shutdown();
+  ctx->config_stream_client->Shutdown();
 #else
   // This code is not unit tested.
 #endif