Simon, can you add the alerts? (Make sure to alert per-page, it's different levels for system_health than for v8)

I think you want this?

General perf sheriff:
ChromiumPerf/*/system_health.common_desktop/total:500ms_window:renderer_eqt_cpu_max/*/*
ChromiumPerf/*/system_health.common_desktop/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/system_health.common_mobile/total:500ms_window:renderer_eqt_cpu_max/*/*
ChromiumPerf/*/system_health.common_mobile/total:500ms_window:renderer_eqt_max/*/*

v8 memory sheriff:
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt_cpu_max/*/*
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt:v8_cpu_max/*/*
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt:v8_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt_cpu_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt:v8_cpu_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt:v8_max/*/*

Not sure I understand what you mean by different levels?

Oops, yeah, actually was incorrect about the levels (I thought it would be a single * at the end for v8 but I was wrong). Patterns in #2 lgtm. Thanks!

Ideally, I don't think we'd alert on both the CPU time and wall clock time variants. We added the CPU time variants to reduce noise.

Somewhat unintuitively, the CPU time metrics are actually noisier (from a relative perspective), likely because the values are much lower overall.

ulan@, do you agree we should only monitor one of these? Which do you think we should monitor?

I'm leaning towards the wall clock time variant, as it appears a bit more stable, on more closely reflects what we measure in the wild.

+1 for monitoring the wall clock time only.

Should we remove the CPU time metric completely?

Wall time are the _max variants?

General perf sheriff:
ChromiumPerf/*/system_health.common_desktop/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/system_health.common_mobile/total:500ms_window:renderer_eqt_max/*/*

v8 memory sheriff:
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/v8.browsing_desktop/total:500ms_window:renderer_eqt:v8_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt_max/*/*
ChromiumPerf/*/v8.browsing_mobile/total:500ms_window:renderer_eqt:v8_max/*/*


If this is all good, I'll go add them!

I'd prefer to leave the CPU variants in place, to compare against when we do see regressions. If it proves useless after a few quarters, we can get rid of it at that point.

That list in #6 LGTM.

> Wall time are the _max variants?
That's correct.

Ok, these have been updated.