Hey there! We've found the following might help you get unblocked faster:

• 📚 Deployment Regions

• 📚 Deployment Teardown

• 📚 Deployment Actions

If you find the answer from one of these, please let us know by solving the thread!

experiencing a ton of little nuances that are stopping us

Hey! Do you mind sharing more details about the issues you're running into?

I ended up figuring out how to deploy. Hoorah! And we just moved our production application here. Woohoo!
So far very happy - much faster, much cheaper.

1. I read MeteorJS applications on Railway cannot have replicas in the exact same region (due to requiring users stickied to a single instance for their session). Do you know anything about that?

2. We have a replica in US West and primary on US East. We're finding 1 gets most of the load, but unable to identify which one.
   So we don't know if we need more US East replicas, or US West, because I can't see a way to filter metrics by replica, etc. Do you have any creative ideas here?

3. If we host a MongoDB with Railway on a DIFFERENT project, is there a way to communicate privately, or will we be charged egres because they're inside different projects and need public communication?

1. You are correct, unless if you are willing to negotiate the websocket state manually to make sure that the session goes to the correct replica go ahead (using something like Caddy) else, the default load balancer wouldn't be a good fit. With that said, you should be fine with a single instance since you can scale to 32 Cores and 32 GBs, something you are not close to hitting in this moment.

2. Related to the above, I am wondering why you are region splitting for now, is there a latency requirement you are trying to work through?

3. Yes, you would need to host the MongoDB instance in the same project to have access to the private network. Curious to why you are looking into hosting it in a separate project?

Thanks for the detailed reply!

1. I think you're right. We should be fine with a single instance.

2. More of a redundancy for high availability. If shit hits the fan for whatever inconceivable reason, it would be nice to know there is an already booted up replica to host the traffic while it comes back to life.

3. We could move it all into a single project, but just feels messy, and the Observability Dashboard doesn't feel granular enough to be confident in monitoring say 15+ different services under the same project. But maybe this the normal way to do it or I'm not versed enough in Railway to understand - definitely not opposed to consolidating it as our current Egress bill is gross.

What are your thoughts?

Ahhh, so like a configurable failover. You could stand up a replicated instance that you can keep around and then switch it to that instance.

For 3. I disagree(?) you can group and organize your services to make it manageable. However, would love your feedback here to see if there are any messy parts to managing your service.

"You could stand up a replicated instance that you can keep around and then switch it to that instance.", is this what we're doing now?

I'm not sure I totally understand what you mean.

For 3, my feedback is:

I don't see any way to filter the usage graphs down to a specific service or replica - it looks like everything is just aggregated at the project level. I know each service has its own Metrics tab, but having to jump between tabs (metrics, then logs, then back to metrics) to correlate different data points is pretty clunky.

What I'd really love is to stay in one dashboard view, like the Observability Dashboard, where I can see how individual services and replicas are performing and pull up the relevant logs for each one without having to bounce around. Would make troubleshooting so much smoother.

In its current state, it feels like for a big project, the OD kind of loses its magic and visibility into the health of everything.

Ultimately though you are totally right. I'll consolidate all this stuff under a single project. No sense paying more for my perceived organization.

If you are referring to this: https://railway.com/project/bfe9f163-53fd-4d50-aa55-b1d03e9037a7? - not so considering I only see one of your application.

On the other note for O11y, you can? When adding additional observability panes, you can add additional filtering to make it so that you can only see the graphs of certain services.

Once you have all of the services in a single place, more than happy to show you the ropes on how to organize everything.