RAPPOR implementation

components/rappor/log_uploader.cc

Index: components/rappor/log_uploader.cc
diff --git a/components/rappor/log_uploader.cc b/components/rappor/log_uploader.cc
new file mode 100644
index 0000000000000000000000000000000000000000..827bd137a677f3580a213db76388873d2e519ca8
--- /dev/null
+++ b/components/rappor/log_uploader.cc
@@ -0,0 +1,134 @@
+// Copyright 2014 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "components/rappor/log_uploader.h"
+
+#include "net/base/load_flags.h"
+
+using base::TimeDelta;
+
+namespace {
+
+// The delay, in seconds, between uploading when there are queued logs to send.
+const int kUnsentLogsIntervalSeconds = 3;
+
+// When uploading metrics to the server fails, we progressively wait longer and
+// longer before sending the next log. This backoff process helps reduce load
+// on a server that is having issues.
+// The following is the multiplier we use to expand that inter-log duration.
+const double kBackoffMultiplier = 1.1;
+
+// The maximum backoff multiplier.
+const int kMaxBackoffMultiplier = 10;
+
+// If an upload fails, and the transmission was over this byte count, then we
+// will discard the log, and not try to retransmit it.
+const size_t kUploadLogAvoidRetransmitSize = 50000;
+
+// The maximum number of unsent logs we will keep.
+// TODO(holte): Limit based on log size instead.
+const size_t kMaxQueuedLogs = 10;
+
+}  // namespace
+
+namespace rappor {
+
+LogUploader::LogUploader(const GURL& server_url,
+                         const std::string& mime_type,
+                         net::URLRequestContextGetter* request_context)
+    : server_url_(server_url),
+      mime_type_(mime_type),
+      request_context_(request_context),
+      has_callback_pending_(false),
+      upload_interval_(TimeDelta::FromSeconds(kUnsentLogsIntervalSeconds)) {
+}
+
+LogUploader::~LogUploader() {}
+
+void LogUploader::QueueLog(const std::string& log) {
+  queued_logs_.push(log);
+  if (!upload_timer_.IsRunning() && !has_callback_pending_)
+    StartScheduledUpload();
+}
+
+void LogUploader::ScheduleNextUpload() {
+  if (upload_timer_.IsRunning() || has_callback_pending_)
+    return;
+
+  upload_timer_.Start(
+      FROM_HERE, upload_interval_, this, &LogUploader::StartScheduledUpload);
+}
+
+void LogUploader::StartScheduledUpload() {
+  has_callback_pending_ = true;
+  current_fetch_.reset(
+      net::URLFetcher::Create(server_url_, net::URLFetcher::POST, this));
+  current_fetch_->SetRequestContext(request_context_.get());
+  current_fetch_->SetUploadData(mime_type_, queued_logs_.front());
+
+  // We already drop cookies server-side, but we might as well strip them out
+  // client-side as well.
+  current_fetch_->SetLoadFlags(net::LOAD_DO_NOT_SAVE_COOKIES |
+                               net::LOAD_DO_NOT_SEND_COOKIES);
+  current_fetch_->Start();
+}
+
+void LogUploader::OnURLFetchComplete(const net::URLFetcher* source) {
+  // We're not allowed to re-use the existing |URLFetcher|s, so free them here.
+  // Note however that |source| is aliased to the fetcher, so we should be
+  // careful not to delete it too early.
+  DCHECK_EQ(current_fetch_.get(), source);
+  scoped_ptr<net::URLFetcher> s(current_fetch_.Pass());

[Alexei Svitkine (slow), 2014/01/24 22:09:03]

Use more meaningful variable names than |s|.

+
+  int response_code = source->GetResponseCode();
+
+  bool upload_succeeded = response_code == 200;
+
+  // Provide boolean for error recovery (allow us to ignore response_code).
+  bool discard_log = false;
+  const size_t log_size = queued_logs_.front().length();
+  if (!upload_succeeded && log_size > kUploadLogAvoidRetransmitSize) {
+    discard_log = true;

[Alexei Svitkine (slow), 2014/01/24 22:09:03]

Add a histogram to log these sort of events, so we can diagnose any issues
through UMA.

+  } else if (queued_logs_.size() > kMaxQueuedLogs) {
+    discard_log = true;
+  } else if (response_code == 400) {
+    // Bad syntax.  Retransmission won't work.
+    discard_log = true;
+  }
+
+  if (upload_succeeded || discard_log)
+    queued_logs_.pop();
+
+  // Error 400 indicates a problem with the log, not with the server, so
+  // don't consider that a sign that the server is in trouble.
+  bool server_is_healthy = upload_succeeded || response_code == 400;
+  UploadFinished(server_is_healthy, !queued_logs_.empty());
+}
+
+void LogUploader::UploadFinished(bool server_is_healthy,
+                                 bool more_logs_remaining) {
+  DCHECK(has_callback_pending_);
+  has_callback_pending_ = false;
+  // If the server is having issues, back off. Otherwise, reset to default.
+  if (!server_is_healthy)
+    BackOffUploadInterval();
+  else
+    upload_interval_ = TimeDelta::FromSeconds(kUnsentLogsIntervalSeconds);
+
+  if (more_logs_remaining)
+    ScheduleNextUpload();
+}
+
+void LogUploader::BackOffUploadInterval() {
+  DCHECK_GT(kBackoffMultiplier, 1.0);
+  upload_interval_ = TimeDelta::FromMicroseconds(static_cast<int64>(
+      kBackoffMultiplier * upload_interval_.InMicroseconds()));
+
+  TimeDelta max_interval = kMaxBackoffMultiplier *
+                           TimeDelta::FromSeconds(kUnsentLogsIntervalSeconds);
+  if (upload_interval_ > max_interval || upload_interval_.InSeconds() < 0)
+    upload_interval_ = max_interval;
+}
+
+}  // namespace rappor