Closed iambriccardo closed 3 months ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 99.01%. Comparing base (
05fa62a
) to head (3c75977
).
:white_check_mark: All tests successful. No failed tests found.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@iambriccardo when deploying this change the relay container simply stops working because it can't connect to the relay container. Isn't the inner relay trying to retrieve the project config from itself now?
@stayallive can you try switching the upstream in the Relay config from:
to upstream: "http://sentry:9000/"
?
Actually nvm, the sentry
is just the Nginx upstream, so should be fine: https://github.com/getsentry/self-hosted/blob/da06c0f230454b91fada3980ba2e1e012cedf1d2/nginx/nginx.conf#L74-L77
Relay should connect to Sentry directly 🤔
I see, my relay is connecting to nginx though... this might have changed in the past... let me correct that!
Okay, user error here. Sorry!
I think a few (hundred) releases ago there was some other issue that was solved by connecting relay through nginx, but that has long been not needed anymore and was forgotten to revert. Seems to all be working now connecting straight to sentry! Thanks for the hint!
This PR updates the
nginx
configuration to forward to inner relays all requests that go to/api/0/relays/*
. The rationale behind this change is that we don't want any more such endpoints to be hit in Sentry (for example when fetching project configs since in https://github.com/getsentry/sentry/pull/72926 we removed the support to return the reduced configuration).