I took a look at the spikes in the GoogleChrome.app graph. Looks like one of the builders is running OS X 10.9.5, and the rest are running OS X 10.11.4. Is it expected to be different with different OS versions? Do you have a preferred version?

In the past, we monitored both GoogleChrome.app and GoogleChromeFramework -- there's some overlap, but having both helps to diagnose regressions. It's perhaps even more useful to have GoogleChrome.app *minus* GoogleChromeFramework for the purpose of generating alerts, but of course it's the total GoogleChrome.app that we most care about so that might confuse people. So... I think we should add alerts for GoogleChromeFramework (and keep GoogleChrome.app with alerts too, but as-is).


I think looks good except for those 200kB spikes. They're weird.

Chrome uses its own compiler pulled in via gclient, and a pinned version of xcode for building the .xib files. There could be very small differences if there are some outdated SDK or standard library files floating around on the system. But I don't think that's the case here. Drilling down - I don't see the spikes in GoogleChromeFramework or GoogleChrome, so it's "something else" inside the .app bundle. Is it easy to grab each .app and diff `ls -lR`?

But also building Chrome on 10.9.5 is basically unsupported - that bot should be upgraded. We target the 10.10 SDK - I think there's a period before the new OS is released where Apple release a beta 10.10 SDK to things running on 10.9. Filed Issue 608233 to get the bots upgraded.


While poking around I noticed some historical data in `sizes->ChromiumChrome->Google Chrome Mac` - did we ever fire alerts for these? (Can we add graph annotations after the fact?)

Link: https://chromeperf.appspot.com/report?sid=929bc8803208e200fc32a2620b9afe26df10045698cb4f76b5679d174668e0ba


And is the difference between `sizes->ChromiumChrome->Google Chrome Mac` and `sizes->ChromiumPerf->mac`:
 - ChromiumPerf is branded, ChromiumChrome is not
 - ChromiumPerf runs every revision, ChromiumChrome might not
 - ChromiumPerf is not accessible without a (@google?) login, ChromiumChrome is public

Are any of those statements false? Are there other differences? Will we keep ChromiumChrome?

>While poking around I noticed some historical data in `sizes->ChromiumChrome->Google Chrome Mac` - did we ever fire alerts for these? (Can we add graph annotations after the fact?)

We did fire alerts starting about 1 month ago. Unfortunately we can't add graph annotations after the fact currently.

>And is the difference between `sizes->ChromiumChrome->Google Chrome Mac` and `sizes->ChromiumPerf->mac`:
> - ChromiumPerf is branded, ChromiumChrome is not

+laforge, phajdan.jr: do you know if ChromiumChrome is official release? Branded?

> - ChromiumPerf runs every revision, ChromiumChrome might not

Yes (except for the blocking bug causing it not to)

> - ChromiumPerf is not accessible without a (@google?) login, ChromiumChrome is public

Whoops, this was an oversight, all should be public now.

> Are any of those statements false? Are there other differences? Will we keep ChromiumChrome?

We'll definitely keep it in the short term, until we're sure nothing is wrong with the ChromiumPerf ones. Do we have any use for it beyond that?

> Do we have any use [ChromiumChrome] it beyond that?

I'm not sure! I'm guessing it's tracking things other than just `sizes`..

(although it would be nice if selecting the `sizes` test suite filtered out bots that have no data for it, or gave some suggestions about where to find historical/current data. E.g. I used to look at  
  sizes->Chromium->chromium-rel-mac
for sizes. It stopped updating in Nov 2015. But goes back to May 2015.
  sizes->ChromiumChrome->Google Chrome Mac
has Sep 2015 to ~now
  sizes->ChromiumPerf->mac
has May 2016 to ~future
  

I'm marking this fixed since the monitoring is updated and the spikes are gone.