One thing we could do is exclude try runs from the counts, that would make
the bugzilla comments less confusing.

As for actual number of runs, we would need to ingest the raw logs in a
reliable more deterministic way- that is a few TB/week.  There are plans to
take ActiveData (which ingests all the raw logs) and put it into BigQuery
which should give us the support we need for a reliable service.  That will
probably be finished near the end of the year.  There are a lot of edge
cases with the current implementation of activedata when we have tests that
don't output .json logs or they don't have unique or useful testnames (like
reftests that output the testname but the test path).

All of these things have been on our plans for a few years, we keep getting
newer higher priority things to do which delays our plans to work on things
like this.  I think once we have a data warehouse in bigquery we should
revisit this topic and look at what remains to be done.


On Mon, Jul 8, 2019 at 2:23 PM Botond Ballo <bba...@mozilla.com> wrote:

> On Mon, Jul 8, 2019 at 1:29 PM Maher( 馬羅成 )Joel <jma...@mozilla.com>
> wrote:
> > Here you can see the pushes we have over time:
> > https://sql.telemetry.mozilla.org/dashboard/firefox-ci
> >
> > looking at a monthly level, I see these numbers for (autoland, inbound,
> central):
> > July 2018 - 2049
> > June 2019 - 2376
>
> Ok, I guess I was misled by the numbers in the "X failures in Y
> pushes" figures in the Intermittent Failures Robot comments. Perhaps
> we could fix those / tweak them to be more reflective of the actual
> failure rate?
>
> Thanks,
> Botond
>
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform

Reply via email to