[features] A health check endpoint

[lzhuor]

Could we please add a simple health api in index.js?

It will serve our needs for load balancers (GCP for us), site reliability monitoring. I am creating a PR for it but seeking for your advice in advance.

[label-actions]

Hey @lzhuor 👋 Friendly reminder: we don't track feature requests on GitHub.

Please look for similar ideas to vote for on the forum and if you can't anything similar then post your own idea.

FYI: Many projects have their own support guidelines and GitHub will highlight them for you as it did here, or the project owners will use issue templates to point you in the right direction. Reading the guidelines or issue templates before opening issues can save you and project maintainers valuable time.

[lzhuor]

Reopening

[jakubgs]

Really? No healthcheck endpoint? Seriously?

[jakubgs]

@lzhuor you said "Reopening" but you never reopened it.

[damusix]

Having issues with a restarting container on ECS. Root gives me a 302 which is considered unhealthy. Ghost needs a healthcheck endpoint to keep up with modern deployment strategies...

[kevinansfield]

If you need a 200 that isn't cached you can hit an API endpoint. /ghost/api/v3/admin/site/ is unauthenticated.

[jakubgs]

 % curl -s 'http://localhost:8083/ghost/api/v3/admin/site/' -H 'X-Forwarded-Proto: https' | jq .
{
  "site": {
    "title": "Example",
    "description": "Example Blog",
    "logo": "/content/images/logo.png",
    "url": "https://blog.exampl.org/",
    "version": "3.27"
  }
}

Looks good. But how could I have possibly found this in the docs?

There should be a /health that shows essentially the same thing and be documented.

[ethankent]

Not having a health check route seems to make Ghost impossible to run behind a load balancer if you need https enabled.

[sneko]

I totally agree, that's a total mess to make such a simple thing. I see 3 things here:

• A dedicated endpoint should exist (since the "solution" above will vary depending on the Ghost version)
• Even if HTTPS is used for the public/admin URL, the server should accept HTTP for the /health endpoint
• Moreover, for the /health endpoint the server should not check the Host: xxxx header because it expect the defined URL so when doing it with localhost:2368/ghost/api/... the connection is refused.

I'm working on a Kubernetes environment, here is an example how you can set the healthchecks on your blog deployment:

        readinessProbe:
          httpGet:
            path: /ghost/api/v4/admin/site/
            port: deployment-port
            httpHeaders:
            - name: X-Forwarded-Proto
              value: https
            - name: Host
              value: blog.mydomain.com
        livenessProbe:
          httpGet:
            path: /ghost/api/v4/admin/site/
            port: deployment-port
            httpHeaders:
            - name: X-Forwarded-Proto
              value: https
            - name: Host
              value: blog.mydomain.com

Hope it helps ❤️

PS: thanks @jakubgs for pointing out the right URL to use 😃

[ErisDS]

The /ghost/api/admin/site/ endpoint is a great workaround for this.

There's a PR open to address this issue here: #13117 but as I've mentioned here, the thing that held me back from merging it is the lack of precedent and it also doesn't seem to really solve the problem above and beyond the /ghost/api/admin/site/ endpoint.

I'm more than happy to merge a PR that has a well thought out and evidenced solution for this :)

[debugger22]

/ghost/api/admin/site/ gives 301 instead of 200 for production.

[JaneJeon]

Of note: with ghost v5, it should be /ghost/api/admin/site, and not /ghost/api/v3/admin/site (or anything that's versioned).

[JaneJeon]

And @debugger22, you can bypass it by passing in the X-Forwarded-Proto: https header like others are mentioning here.