[Offline Pages] Handle prerender FinalStatus UnsupportedScheme as a cancel rather than failure.

components/offline_pages/background/offliner_policy.h

Index: components/offline_pages/background/offliner_policy.h
diff --git a/components/offline_pages/background/offliner_policy.h b/components/offline_pages/background/offliner_policy.h
index d640d2590a8cfae7411d8149ecbd3429f2e6c3b5..d3351c82991d002d3972c545f6947f96c575462d 100644
--- a/components/offline_pages/background/offliner_policy.h
+++ b/components/offline_pages/background/offliner_policy.h
@@ -6,7 +6,7 @@
 #define COMPONENTS_OFFLINE_PAGES_BACKGROUND_OFFLINER_POLICY_H_
 
 namespace {
-const int kMaxStartedTries = 5;
+const int kMaxStartedTries = 3;

[dougarnett, 2016/08/30 20:28:15]

Per dimich request to reduce tries from 5 to 3. On GIN-2g I had a few pages take
3 tries to succesfully load (2 x 2 minutes deadbeat timeouts and then success)
so I would not want to go lower than 3 tries.

[Pete Williamson, 2016/08/30 20:42:47]

This might be better as an unrelated changelist.

[Dmitry Titov, 2016/08/30 23:15:38]

Note 5 or 3 are kinda theoretical w/o actual measurement. I meant that in
general if we use simple heuristics and retry on potentially wide bucket of
outcomes, some of which could be permanent in nature, it makes sense to limit
the number of retries. I don't have specific pref to a particular number.

Do we have enough UMA to see outcomes like "took N retries to load"?

[dougarnett, 2016/08/30 23:27:47]

We don't yet have that UMA and agree we need it. I could add as part of this CL
tomorrow (or as separate CL).

[dougarnett, 2016/08/31 16:10:52]

Per Pete's original request, reverting this change from this CL and will add UMA
in follow-up CL (in which can decide whether to speculatively adjust this
lower). I would like to try some of the EM urls on GIN-2gpoor to help decide
that.

 const int kMaxCompletedTries = 1;
 const int kBackgroundProcessingTimeBudgetSeconds = 170;
 const int kSinglePageTimeLimitSeconds = 120;