probe-scraper
probe-scraper copied to clipboard
mozilla
How can I include Glean SDK definitions files that do not reside in Git?
The FOGotype (Firefox on Glean Prototype) integration will be sending an empty "prototype" ping every so often. For that to not be rejected by the pipeline, I gather that the Probe Info Service needs to know about it.
From the docs I gather I need to point the scraper at a Git repo that contains a pings.yaml... I think.
Do I need to provide a repo if there isn't a metrics.yaml? What if it's a mercurial repo instead of Git? Does probe-scraper know about pings.yaml files?
@chutten just for clarity: you should not change the files yourself, in general (as a consumer of Glean). You should rather follow the instructions in the docs.
@Dexterp37 Ah, so making probe-scraper compliant to the particular characteristics of my project is a subordinate task of the Data Engineering bug I filed over here?
@Dexterp37 Ah, so making probe-scraper compliant to the particular characteristics of my project is a subordinate task of the Data Engineering bug I filed over here?
As a user, yes, you should not need to care about probe-scraper at all.
Aside for that, mozilla-central has a Git mirror that's being used for engine-gecko already, so it might be helpful for the FOGotype as well :)
Do I need to provide a repo if there isn't a metrics.yaml? What if it's a mercurial repo instead of Git? Does probe-scraper know about pings.yaml files?
We can provide an empty repository without a metrics.yaml, though we would prefer that there be one available. The pings will contain just the default glean metrics.
Aside for that, mozilla-central has a Git mirror that's being used for engine-gecko already, so it might be helpful for the FOGotype as well :)
This is how I would hope to do it, when the metrics.yaml is available.
Does this all sound reasonable to you @chutten?
There'll be a custom ping coming from the FOGotype ("baseline" pings don't have a lifecycle analogue on Desktop, "metrics" pings would require a lot of wall clock scheduling complexity that we don't want in the prototype, "events" pings would probably work but then we'd need to support glean_parser and be able to generate metrics APIs): the "prototype" ping. It'll be an empty ping. Will that be problematic, @fbertsch ?