self-hosted
self-hosted copied to clipboard
getsentry
Self-hosted beacon probably not working
There are currently <20 self-hosted beacon responses recorded in our internal metrics for all Sentry 20.x versions. This is orders of magnitude smaller than the numbers we're getting for Sentry 9.x and 10.x. This seems incorrect.
A quick glance at GitHub issues filed by users indicates that 20.x issues make up at least 75% of issues filed, so either 20.x users are dramatically more likely to file GH issues and/or dramatically less likely to enable the beacon, or the pipeline is broken in later versions.
Looking into this, I believe the beacon might still be giving us good data. It even shows us our dogfood instance. At least here: https://sentry.io/_admin/beacons
Perhaps the looker dashboard to track all the instances of self-hosted is broken
It seems like whatever appears in the admin dashboard is based on the install id of a self-hosted instance, which is treated as unique. This value is a sha1 hash
I believe we can rely on BigQuery as a source of truth for beacon_checkin, beacon_install, and beacon_stats
Something is not right for Events(24H), even our instance shows this as 0 events, which is not true. The recent beacon checkins with versions 23.* all show up with 0 events there.
This line is returning 0: https://github.com/getsentry/sentry/blob/37e47a02a7a0663315d0be4e69d93224ad434995/src/sentry/tasks/beacon.py#L72-L74
That doesn't seem to be correct. Seems to be from using TSDBModel.internal. Since using TSDBModel.project gives me an accurate event count
Talked about not storing ip addresses, but having a general idea of where people are using self-hosted Sentry would be useful