Making Supabase work in production

[TueeNguyen]

Kong perhaps doesn't play right with our nginx which make our parser service in production can't interact with Supabase database.

Seeding feeds into Prod Supabase also results in the same error An invalid response was received from the upstream server

Related issue Kong/kong#3314

{"level":30,"time":1650669868347,"pid":8,"hostname":"35bbf749cafa","msg":"Loading all feeds into feed queue for processing"}
{"level":50,"time":1650669868463,"pid":8,"hostname":"35bbf749cafa","error":{"message":"An invalid response was received from the upstream server"}}
{"level":50,"time":1650669868464,"pid":8,"hostname":"35bbf749cafa","err":{"type":"Error","message":"An invalid response was received from the upstream server","stack":"Error: An invalid response was received from the upstream server\n    at getAllFeeds (/app/src/utils/supabase.js:23:13)\n    at processTicksAndRejections (node:internal/process/task_queues:96:5)\n    at async processAllFeeds (/app/src/parser.js:62:19)"},"msg":"Error queuing feeds"}

[DukeManh]

Kong didn't work because it made requests to supabase-rest which failed to connect to the DB using the connection string.

I notice the Postgres password has a "@" in it, and that's a problem because the connection string is in the form postgres://user:pass@host, notice the "@" in the middle. If the password contains a "@", we'd have to escape it or use a different password.

I updated the password on the production run migration and data seeding, and it seems to be working.

@TueeNguyen, please test the production supabase and let me know
@humphd, I just used a uuid for the password, maybe you want to use a different one, it doesn't have to be a JWT.

[humphd]

Great detective work! I've made updates to the values we put in the secret envs, and sent to @manekenpix to be set on staging and production. Once he does that, we'll need to do new builds (e.g., push something to master and deploy a new tag) for them to get picked up properly. Hopefully we can pair this with landing @DukeManh's work in #3558.

[TueeNguyen]

Seems like it's working on production now with the new password

[humphd]

@TueeNguyen only temporarily. A push/tag will re-create the old env vars. We need it fixed in the secrets env files for this to stick.

[TueeNguyen]

Right, I just looked at Slack's convo

[humphd]

@manekenpix reports that this is now done on staging/production. Hurray!

We need a push/tag to update staging/production to use these, and to confirm the fix.

[humphd]

I'm going to close, and we can do follow-ups as we discover new problems.