This is the project page for the machine vision extension: https://www.mediawiki.org/wiki/Extension:MachineVision
Developer docs: https://www.mediawiki.org/wiki/Extension:MachineVision/Developers
Repository: https://gerrit.wikimedia.org/g/mediawiki/extensions/MachineVision

Adding my manager @ssastry for visibility

1. type is not considered metadata in our message definition. Currently metadata is only the topic.
2. What kind of logic do you expect on the app side? I guess having a new checkEchoV2 depends in future product needs. Also usage is pretty low at the moment so there is space to adapt to client needs.

Fix is in prod:

1. In the service layer we do batching based on the combination of:
   • notification token
   • type
   • provider
   • message metadata.

Closing this one after: https://www.mediawiki.org/wiki/Technical_Decision_Making_Process/Technical_Forum/Published_Proposals/T275063

It looks like there is already some progress on providing bullseye docker images.
https://phabricator.wikimedia.org/T281596

Here is a report of how tegola performs to various scenarios compared with the current kartotherian setup.
https://gerrit.wikimedia.org/r/c/maps/spikes/+/683824

It looks like native prometheus metrics are now exposed in the service. That said we may still need to adapt the grafana dashboard because the metrics names might have changed.

Metrics still not working. Things i discovered while debugging:

Maybe that's relevant:
(from the logs)

endTiming() unsupported for metric type Gauge

Got it, thanks @Tsevener i am gonna take a look now.

I took a look at the pcs level:

I think this is the patch that introduced the change from statsd metrics to native prometheus:
https://gerrit.wikimedia.org/r/c/mediawiki/services/chromium-render/+/558213

I think the only issue we have from upstream at the moment is: https://github.com/go-spatial/tegola/issues/714
I expect this to be broken down to more atomic issues, but I think we can use this one to track the work happening

If the messages are actually targeting production apps (testflight/app store) wouldn't it make sense to use the production deployment directly?

Hey @fgiunchedi, thanks for the feedback. I think what we are trying to show in the charts is the differentiation between "read-only" and "read-write" access. I will update the chart so we can address the topic you raised.

Closing this one after: https://github.com/wikimedia/kartodock/commit/0c581f5fc8c5f6ded206994337a37b714164433f

It looks like tegola can be used with Swift using s3api combatibility.
Local setup:

@MSantos the memory leak fix is in production for some time now, should we close this one?