Add a histogram to log the size of an UMA upload.

components/metrics/metrics_service.cc

Index: components/metrics/metrics_service.cc
diff --git a/components/metrics/metrics_service.cc b/components/metrics/metrics_service.cc
index ca9fef4fea954d01245ca6370557befef2464fa7..fe341639d61dedc444d4d6babf4eb8025b29696c 100644
--- a/components/metrics/metrics_service.cc
+++ b/components/metrics/metrics_service.cc
@@ -1026,7 +1026,9 @@ void MetricsService::OnLogUploadComplete(int response_code) {
   // Provide boolean for error recovery (allow us to ignore response_code).
   bool discard_log = false;
   const size_t log_size = log_manager_.staged_log().length();
-  if (!upload_succeeded && log_size > kUploadLogAvoidRetransmitSize) {
+  if (upload_succeeded) {
+    UMA_HISTOGRAM_COUNTS_10000("UMA.LogSize", log_size / 1024);

[Ilya Sherman, 2014/08/30 01:23:23]

Hmm, why not record this for all logs when sending the upload, rather than only
for successful logs after the upload response is received?  I guess it's not
obvious to me which is more useful; but the filtering by success does seem like
it could limit visibility into any potential problems with discarding large
logs.

[Alexei Svitkine (slow), 2014/09/02 14:50:37]

I want to use this histogram to capture how much data is sent up to Google
servers (and do so client-side so that e.g. I can see the effect of experiments
on this - e.g. playing with upload frequency), so I'm only interested in
successes (to avoid e.g. double counting on re-transmissions).

+  } else if (log_size > kUploadLogAvoidRetransmitSize) {
     UMA_HISTOGRAM_COUNTS("UMA.Large Rejected Log was Discarded",
                          static_cast<int>(log_size));
     discard_log = true;