Bugzilla

(In reply to Robin Steuber (they/them) [:bytesized] from comment #4)
> (In reply to Chris H-C :chutten from comment #3)
> > With present technology I'd recommend having a second event with a similar name (maybe with a suffix like `_ext`) with the extra extras. There's a possibility we might only receive one or the other event instead of both, but that should only happen if the "event" ping splits between them and one of the two pings fails to be received properly (ie, should basically never happen).
> 
> @nalexander does this sound like an improvement over the current situation? Should I get started on this?

Argh -- a partially typed comment that I missed finishing.  Anyway:

Thanks, :chutten.

:bytesized: yes, I think we'd better to do this, since I'm not having a lot of luck joining `installation.first_seen` events to `new-profile` pings in *either* direction.

Earlier I witnessed the reverse, lots (>50%) of `new_profile` pings without `installation.first_seen` events, over in https://bugzilla.mozilla.org/show_bug.cgi?id=1811374.  (That is expected, prompting that ticket.)  See https://sql.telemetry.mozilla.org/queries/90885/source?p_range=2022-12-01--2022-12-28.

But now I witness lots of `installation.first_seen` events without corresponding `new_profile` pings, perhaps >60%!  See https://sql.telemetry.mozilla.org/queries/90886/source?p_range=2022-11-01--2022-11-29.

So yes, we'd better get an event at all, in the hopes that we'll get two events and be able to actually have coherent data from the single system.  Thanks!

(In reply to Robin Steuber (they/them) [:bytesized] from comment #4)
> (In reply to Chris H-C :chutten from comment #3)
> > With present technology I'd recommend having a second event with a similar name (maybe with a suffix like `_ext`) with the extra extras. There's a possibility we might only receive one or the other event instead of both, but that should only happen if the "event" ping splits between them and one of the two pings fails to be received properly (ie, should basically never happen).
> 
> @nalexander does this sound like an improvement over the current situation? Should I get started on this?

Argh -- a partially typed comment that I missed finishing.  Anyway:

Thanks, :chutten.

:bytesized: yes, I think we'd better to do this, since I'm not having a lot of luck joining `installation.first_seen` events to `new-profile` pings in *either* direction.

Earlier I witnessed the reverse, lots (>50%) of `new_profile` pings without `installation.first_seen` events, over in https://bugzilla.mozilla.org/show_bug.cgi?id=1811374.  (That is expected, prompting that ticket.)  See https://sql.telemetry.mozilla.org/queries/90885/source?p_range=2022-12-01--2022-12-28.

But now I witness lots of `installation.first_seen` events without corresponding `new_profile` pings, perhaps >60%!  See https://sql.telemetry.mozilla.org/queries/90886/source?p_range=2022-11-01--2022-11-29.

So yes, we'd better get an additional event, in the hopes that we'll get two events and be able to actually have coherent data from the single system.  Thanks!