add a healthcheck to the companion

[trashhalo]

This will let me set a readiness probe for the pods

[github-actions]

Diff output files

diff --git a/packages/@uppy/companion/lib/companion.js b/packages/@uppy/companion/lib/companion.js
index 055c7ec..5da3e96 100644
--- a/packages/@uppy/companion/lib/companion.js
+++ b/packages/@uppy/companion/lib/companion.js
@@ -84,6 +84,7 @@ module.exports.app = (optionsArg = {}) => {
     app.use(middlewares.metrics({ path: options.server.path }));
   }
   app.use(cookieParser()); // server tokens are added to cookies
+  app.get("/health", (req, res) => res.send("OK"));
   app.use(interceptGrantErrorResponse);
   // override provider credentials at request time
   // Making `POST` request to the `/connect/:provider/:override?` route requires a form body parser middleware:

[kvz]

What about using the /metrics endpoint for this?

[trashhalo]

What about using the /metrics endpoint for this?

I disabled /metrics because I didn't want server metrics to be publicly available.

[kvz]

It's a better health indicator imho tho because

• more things need to work in order for it to give a 200
• you ensure that observability works if /metrics is also your health check

[trashhalo]

Okay do you want me to remove the new endpoint and just update the kube docs to use /metrics as the readiness probe?

[kvz]

Yeah that would be good!

[mifi]

maybe in the future we want to send some json data here. so users should not expect an "OK" text here.

Suggested change

	app.get('/health', (req, res) => res.send('OK'))
	app.get('/health', (req, res) => res.end())

[kvz]

Was leaning towards just using /metrics for this Mikael, it’s a more trustworthy check, end we don’t need to introduce a second health-y endoint then

[mifi]

that sounds good to me