MultiConnectMultiConnectDocs Website home
MultiConnectCloud › Server Deployments

3. Create a server deployment

  1. On the dashboard, open your project.
  2. Click Create New Deployment, then choose Server.
  3. Work through the steps below. Next checks each step before you move on and lists every problem at once. Back returns to an earlier step.
  4. On the last step, click Submit. Your build uploads from the browser, and the deployment starts.

The Deployment guide button at the top of every step opens the expected build layout and integration notes for the platform you picked.

Step 1: Basic Info

Field What to enter
Deployment Name 3 to 60 characters: letters, numbers, spaces, hyphens and underscores, starting with a letter or number. It must be unique in the project. Example: arena-servers
SDK App Optional. The game (SDK app) these servers host, if it uses MultiConnectSDK. Its servers get the app's SDK key from their first start, and players whose game uses MultiConnectCloud with that key are sent here. An app already linked to another deployment is listed but cannot be chosen. Create apps on SDK Apps

When an SDK app is linked, your server build can leave the SDK key empty: on MultiConnectCloud it is read from MC_SDK_KEY. Game clients still need the key in their build.

A project can have up to 20 deployments.

Step 2: Mode & Plans

Field What to choose
Mode Combined or Master–Worker. See Section 2
Plan (Combined) or Worker / server plan (Master–Worker) The CPU and memory each server or worker container gets, and how many containers a region can run. You pay the plan rate for every running container, by the hour
Master plan (Master–Worker) The plan for master containers. Masters usually do far less work than the servers running matches, so a smaller plan is often enough
Game servers (Master–Worker) Always ready or On demand
Most running at once (On demand) Your cost limit: the most game servers that may run at once, from 1 to 50. When all are in use, new players wait or are told to try again
Longest a server may run (minutes) (On demand) A safety limit, from 10 to 240. A game server still running after this is stopped

Plans with no capacity available are shown but cannot be chosen.

Step 3: Runtime

Field What to enter
Platform The runtime your server runs on. See Section 4. Choosing one fills in its default entry point
Entry Point File The file that starts your server, at the top level of your zip. Unity: the server binary, such as MyGameServer.x86_64. It is required, because a Unity binary is named after your project. .NET: your DLL. Node.js: usually blank
Start Command Leave it empty to use the platform default, which the form shows under the field. Fill it in only if your server must start differently, such as node server.js. One program with its arguments: the form refuses shell features such as && or | on this step
Container Port The port your server listens on inside the container, on 0.0.0.0, from 1024 to 65535. Each container gets its own public port, which your process can read from MC_PUBLIC_PORT
Protocols What the port accepts. Tick every one your game uses (see below)
Health Check Path (optional) An HTTP path that answers 2xx once your server is ready, such as /healthz
Environment Variables Settings your server reads at start-up (see below)
Grace Period and Scale-Down Grace Period See Section 8

Protocols. All protocols use the same container port.

Protocol Use it for
TCP Reliable connections: socket servers and turn-based games, with native clients
UDP Fast real-time game traffic with native clients. WebGL clients use WebRTC on the same port
WebSocket A secure wss:// address for each container, needed for browser (WebGL) clients. It runs over TCP, so ticking it also includes TCP, and the port also accepts plain TCP

Common choices: UDP for action games, WebSocket and UDP for games with both browser and native clients, TCP and UDP for a reliable channel plus a fast one. At least one of TCP or UDP stays selected. A UDP-only server has no wss:// address.

Health check. Health decides when a container is announced to players, and when a redeploy moves new players to the new servers. If you set a path, the platform sends an HTTP GET to it. Without one, a TCP server is checked by connecting to its port, and a UDP-only server counts as healthy once it has stayed up for a while. Setting a path is worth it.

Environment variables. Click Add Environment Variable for each one, or Upload .env File to read KEY=VALUE lines from a file (existing keys are updated). Set Secret to Yes to store a value encrypted and hide it after saving.

Rule Detail
Names Letters, numbers and underscores, starting with a letter or underscore
Reserved Names starting with MC_ are set by the platform and cannot be used. Click Environment reference on the form to see them, or read Section 10
Duplicates Each name once

Step 4: Topology

Combined mode. Tick every region you want a server in, and click the star on one to make it the primary. Each ticked region runs exactly one container.

Master–Worker mode.

  1. Under Master servers, tick each region that should run a master, and star one as primary. Each master region runs exactly one master.
  2. Under Worker servers, click Add Worker Region for each region workers should run in, and fill in its row:
Column Always ready On demand
Region Where these workers run Where game servers may start
Min Workers always running, and always billed. At least 1 Not shown. Always 0, because nothing runs until your master asks
Max or Most at once How far autoscaling may grow, up to the plan's limit per region The most game servers that may run in this region at once, up to the plan's limit per region. A cap, not a target
Reports to The master region these workers report to The master region these game servers report to

You need at least one master region and at least one worker region. A region the chosen plan cannot use is shown disabled, with the reason. Master regions are fixed after create: on Update they are shown read-only.

Step 5: Autoscaling

This step appears only for Master–Worker deployments with Always ready game servers. Combined servers run one container per region, and on-demand game servers are started by your master, so neither autoscales.

Field Meaning
Auto Scaling Automatic (ON): workers are added and removed with load. Managed (OFF): each region runs the count you set
CPU Scale Up Threshold Average CPU use (%) above which a worker is added, for example 80
CPU Scale Down Threshold Average CPU use (%) below which a worker is removed, for example 30. It must be below the scale-up value. Keep a clear gap
Memory Scale Up Threshold Average memory use (%) above which a worker is added
Memory Scale Down Threshold Average memory use (%) below which a worker is removed. It must be below the scale-up value

Autoscaling always stays within each region's Min and Max.

Step 6: Webhook

Turn on Webhook and enter an https:// URL, such as https://example.com/webhooks/multiconnect. The platform then sends signed events to it when servers are added, retiring, removed or unhealthy, and when a deploy fails or the deployment is stopped or deleted. In Master–Worker mode without MultiConnectSDK, your master needs this to learn when workers appear and disappear. See Section 11.

Step 7: Build

Click Choose zip and pick one .zip of your built server. Master–Worker deployments need two: Master build and Worker build. The entry point must be at the top level of the zip, not inside a folder. The size limit comes from your worker or server plan, and applies to both builds.

Your balance must cover at least one hour of the deployment at its minimum container counts, or the form says so before anything is uploaded.

After Submit, the deployment page opens and shows the operation's progress. The first containers come up, pass their health check, and the deployment is live.