Report the running version as a metric

[DMRobertson]

Allows one to display (or superimpose) the version onto grafana graphs, to cross-reference with deployments.

[DMRobertson]

These should be underscores. If not, runtime error.

Need to enable prometheus metrics endpoint in the E2E tests.

[kegsay]

We have an entire file called metrics_test.go in integration which do this.

[kegsay]

Ah but it hooks in further down the stack when making the handlers, not in the cmd binary.

[kegsay]

This could be as trivial as starting the binary with the SYNCV3_PROM env var set then ensuring /metrics 200 OKs.

[kegsay]

Pending E2E tests to check it works.

[kegsay]

This isn't what I was expecting. I would rather we didn't bundle this with end-to-end tests, as it just complicates things.

I was expecting more a simple metrics job which built, ran and curled /metrics to verify that it returns some data.

[kegsay]

For example, we currently have:

            - name: Check /client works
              run: ./tests-e2e/client-check.sh
              env:
                  SYNCV3_DB: user=postgres dbname=syncv3 sslmode=disable password=postgres host=localhost
                  SYNCV3_SERVER: https://matrix-client.matrix.org
                  SYNCV3_SECRET: itsasecret

where client-check.sh is:

#!/bin/bash -eu
export SYNCV3_BINDADDR=0.0.0.0:8844

# Run the binary and stop it afterwards.
./syncv3 &
SYNCV3_PID=$!
trap "kill $SYNCV3_PID" EXIT

# wait for the server to be listening, we want this endpoint to 404 instead of connrefused
until [ \
  "$(curl -s -w '%{http_code}' -o /dev/null "http://localhost:8844/idonotexist")" \
  -eq 404 ]
do
  echo 'Waiting for server to start...'
  sleep 1
done

echo 'Checking client is reachable...'
curl -f -LI "http://localhost:8844/client/"

We could easily extend this to check /metrics.

[DMRobertson]

Oh, I've only just noticed the word

extend

in your comment. I can make this one script if you'd prefer.

[DMRobertson]

Gonna drop this for now.