Server Management

Looking for input on server management.

I am running into server sprawl real quick and would like to know what others are doing.

I have barely started building my servers and already have the need for:

Web Server
Vault server
World Server
Advertisement Server
SimEthereal Server
Vault Agent for each SimEthereal Server.

Not to mention I will be adding player stats server and maybe some others like a database server. Potentially 13 servers spread across 10 vps.

I have no problem with the coding/patching part, its the Startup, shutdown, notification if one crashes, that kind of thing I am talking about.

3 of the 6 will be on one VPS/dedicated, so thats 4 different VPS that needs to be managed.

Has anyone experience with this type of thing?

For starters, auto-restart scripts are your friend here - schedule them with cron (assuming Linux) and have them check the process names/ids to make sure the right processes are up - that’s a decent first step if a process crashes. If you’re game to incorporate another server, the Elastic stack has a lot of tools for recording logs and monitoring your servers (there’s also Prometheus, which I’ve never used - I’d personally lean towards Elastic).

If you want to go crazy, you can even do a Docker or Kubernetes cluster but I wouldn’t recommend doing that - those tend to be a huge pain and have such a list of caveats that it’s frankly rarely worth it unless you’ve got a large operation and staff to handle it. Containers are a pain to deal with even when not in a cluster, and everything gets harder as soon as you move to a clustered environment.

Edit: Alerting isn’t built into the core Elastic systems - it’s possible to configure email alerts as part of Logstash, however. You might also want to go a little simpler with something like Nagios Core - here’s a setup for configuring it to send emails (https://www.unixmen.com/send-alerts-nagios-core-using-gmail-yahoo/).

Personally, I’d probably start with auto-restart scripts + periodic, regular email health updates & immediate email alerts when something goes wrong. If you want to get fancy you can even use a service like Twilio to send you an SMS, though you’ll probably need to do some scripting to pull that off.

Docker, it supports healthchecks and can restart the containers automatically if they crash ( --restart=always )

The most robust solution is to develop a backend app that can start and stop the servers and display the information you require to administer them.

The simplest solution is a pinger. It just pings adresses/ports and tells you if it’s down.

You could start with a combination of the two and build upon it as time goes by. A simple web page with a start stop and ping button, and as time goes by make a page for each server type.

1 Like

I come from a background of dealing with large JavaEE applications and large server deployments, so I tend to stick to technology that I am familiar with, but I use docker for my server applications to keep my deployments clean, and use elastic scaling on my clusters. I deploy my apps in AWS.
I build all of my apps for my JME servers as WildFly applications and manage the cluster with a custom JEE management platform I built. This keeps my technology consistent as I am also using Vue on JEE for my websites. This way I only need two different services to worry about: WildFly, and PostgreSQL (for my databases). My JME app uses the pgjdbc-ng driver to communicate with the databases via a connection pool to help reduce the number of connections being created and destroyed between the JME cluster and my DB cluster.