Meteor deployment for testing/production

I have a meteor app for the Indian School ecosystem, it essentially helps in catering to all interactions,information and communication that a parent or a student needs to have with the school. It has a small user/school base. (3000 - 5000 active users daily) We have a two server setup - one for database and another for the rest. And I use S3 for storing images and videos, and DO Spaces for backups (twice a day of DB)

Till last year I was happy using digital ocean using mostly smaller machines. I am now trying to explore moving things to using Kubernetes or docker swarm(not sure how long it will be supported)

Now my question is what would be the minimum cost cheapest(dollar) for setting up the most simple configuration for a meteor app in production in Kubernetes. It would be great if the community can recommend some providers and their experience with them. I would like to test and then possible move to it if needed. Thanks

I tried using minikube but found it extremely (may be it is my lack of knowledge) difficult to setup a mock test setup of production on local machine. (most of the documents are for simple node app setup) Setting up storage and loading data to mongodb was extremely difficult(time out loading larger datasets ). And found difficult to reference pods inside settings - specially redis – but eventually was able to do it. And setting up controllers for load balancing…

Do you really need k8s?

We use Ansible with 50loc to deploy docker container with our app. It is simple and elegant.

Now we are using DO, but will move to Hetzner(bare metal), it is cheapest provider for us with good quality.

I think k8s for huge apps with tens/hundreds services.

1 Like

Agreed. Kubernetes may be an overkill and with simple logic it looks like a minimum setup with kubernetes may be anywhere upwards of $50, but I wanted to know others experience.

Wanted to have options and probably have the experience of setting it up

With baremetal or vps the problem I notice, one has to always provision keeping in mind the possible peak potential and this results in an underutilised resource most of the time.

We had very good experience with Hetzner in terms of reliability and customer service. We just moved away since we now use our corporate cloud, otherwise we would have stayed.

1 Like

Yea, we have the similar issue, we also have online school and our resources stay free by night.

But, for ~50$/month you can buy 64GB RAM, 512GB storage, AMD Ryzen 5 3600 Hexa-Core and use it for apps, redis, monitoring, logs and other stuff(running in separate docker containers).

Anyway, with DO you haven’t auto scaling(should be implemented manually). I know Galaxy do it well, but it costs ~x2-x3 times more than Hetzner/DO.

Do you have your own DC? What about HA?

Yes we have our own (university) and we do not distribute our apps worldwide so we are good with it.

1 Like