#Questions about the service

10 messages · Page 1 of 1 (latest)

marsh geode
#

I am the founder of a startup with quite a bit of users/traffic and I am considering moving to railway on the team plan, but I have a few questions.

I don't really understand how railway internally manages my apps, and I'm struggling to understand the CPU model.

1.) On the developer plan is says 1 shared CPU while on the team it says 32 vCPUs per container. Do you differentiate between CPU and vCPU?

2.) If I go into my account I can see the developer goes up to 8 vCPUs per container. Is this an error in the pricing page?

3.) What does up to mean? what does up to 32 vCPUs per container mean? Does that mean each container runs exactly 1 instance of it and is guaranteed to get 32 vCPUs at any time or does it spawn multiple instances depending on the load, potentially on multiple nodes?

4.) Is my app highly available?

5.) Is there a way to select the DC where to host my app?

6.) I need to host a postgres database with a quite a bit of traffic on it but I heard that your database are not very reliable at the moment, is this true?

image.png image.png
vital pivotBOT
#

Project ID: N/A

marsh geode
#

N/A

sullen arrow
#

Answering to the best of my ability, i’ll tag in the team for the rest.

  1. Still unsure why it says 1CPU on dev plan.

  2. Dev plan provides 8GB RAM and 8vCPU per container.

  3. Your app will scale vertically if needed. Your first point is correct. Railway does not scale horizontally.

  4. Not sure what you mean by this. Once hosted, your app will be available 24/7. The team is currently working on deployments to multiple regions. Atm Railway is limited to GCP-US-West1, though users can connect from anywhere.

  5. Reliable, yes. There are some things that could be improved. General recommendation is to host your db elsewhere.

#

@marsh geode

vital pivotBOT
#

Flagging this thread. A team member will be with you shortly.

shrewd cosmosBOT
#

Hey there Davide,

Angelo here from the Support tool, don't be alarmed.

Your workloads run in a container with CPU and Memory limits. Your application gets to use resources up to those limits. The dev plan is up to 8 vCPU and 8 GB per container. We don't automatically spin up new containers as your project increases load. (WIth that said, we expect HA to come to our beta program soon.) As mentioned, US-East only, but we are working on adding additional regions.

Re: databases, I'd say we are better on our reliability story but admit that we have had some difficulties in the past. We did some work that, knock on wood, have neutralized these issues in the last two weeks. Actions speak louder than code, in our 3 years of operation, we have had no major data losses on behalf of users. (Except twice, where we recovered the users data back in both cases.)

Let us know if you have anymore questions!

marsh geode
#

Does my container get replaced during resize?

sullen arrow
#

or hit the restart button