New: HTTPS on your domainSee the settings

Zero-downtime deploys for one VPS

Sling builds one exact commit, waits for 3 healthy checks before traffic moves, then rolls back by itself if it fails.

Every deploy on record

Each stage of every deploy and rollback is saved to SQLite, so the history survives a restart. A deploy killed halfway is marked interrupted, with the stage it reached. Read the history docs

sling / history

D9c41d0edeploy · Nov 08 16:52health_checking22s
D53a2b2fdeploy · Nov 08 16:46live34s
D399076adeploy · Nov 08 16:46interrupted2s
R67dd979rollback · Nov 08 16:45live33s
Dfeb2e72deploy · Nov 08 16:45failed2s
Da07c3e5deploy · Nov 08 15:12rolled_back41s
D67dd979deploy · Nov 08 14:30live35s
D5d02b6fdeploy · Nov 08 14:02superseded0s
De3b8f41deploy · Nov 08 13:58live36s
De3b8f41deploy · Nov 08 13:57skipped0s
D2c9a7d0deploy · Nov 07 18:20live34s

you/app/maindeploy21.7shealth_checking

MBTime out slow health checks9c41d0e

Clonegit checkout --detach1.4s
Builddocker build19.2s
ProbeGET /health1.1s
SwitchPOST /loadQueued
Watch30s graceQueued

$git clone --no-checkout https://github.com/you/app /tmp/paas-builds/9c41d0eCloning into '/tmp/paas-builds/9c41d0e'...remote: Enumerating objects: 412, done.remote: Counting objects: 100% (412/412), done.remote: Compressing objects: 100% (231/231), done.Receiving objects: 100% (412/412), 188.20 KiB, done.Resolving deltas: 100% (164/164), done.$git checkout --detach 9c41d0eHEAD is now at 9c41d0e Time out slow health checks$git rev-parse HEAD9c41d0e matches the commit asked for$docker build -t paas-app:9c41d0e /tmp/paas-builds/9c41d0e#1 [internal] load build definition from Dockerfile#2 [internal] load metadata for docker.io/library/node:22-alpine#3 [1/5] FROM docker.io/library/node:22-alpine#4 [2/5] WORKDIR /app#5 [3/5] COPY package*.json ./#6 [4/5] RUN npm ci --omit=dev#7 [5/5] COPY . .#8 exporting to image#8 naming to docker.io/library/paas-app:9c41d0e doneBuilt paas-app:9c41d0e$docker run -d --name app-9c41d0e --network sling-net -p 127.0.0.1::3000 --memory 256m --cpus 0.5 -e APP_VERSION=9c41d0e paas-app:9c41d0e$docker port app-9c41d0e 3000127.0.0.1:49731Waiting for http://127.0.0.1:49731/health to pass health checks...+0.5s container running, GET /health: connection refused+1.0s container running, GET /health: connection refused+1.5s container running, GET /health: connection refused+2.0s container running, GET /health: 503 Service Unavailable+2.5s container running, GET /health: 503 Service Unavailable+3.0s container running, GET /health: 200 OK (1 of 3)+3.5s container running, GET /health: 200 OK (2 of 3)needs 3 in a row before 30s$curl -X POST localhost:2019/load (upstream app-9c41d0e:3000)Switching traffic: "app-53a2b2f:3000" -> "app-9c41d0e:3000"Verifying through the proxy (http://127.0.0.1:8080)...GET /health 200, X-Sling-Upstream: app-9c41d0e:3000 (1 of 3)GET /health 200, X-Sling-Upstream: app-9c41d0e:3000 (2 of 3)GET /health 200, X-Sling-Upstream: app-9c41d0e:3000 (3 of 3)The new version is the one answeringObserving for 30s before retiring the previous version...probes through the proxy, rolls back on 3 failures in a row+10s every probe passed+20s every probe passed+30s grace window passedRetiring previous version app-53a2b2f (SIGTERM, up to 10s to drain)...app-53a2b2f stopped, image paas-app:53a2b2f kept for rollbackLive: 9c41d0e

Traffic moves last.

Sling points Caddy at the new container only after it answers /health three times in a row. Then it checks again through the proxy, as your users reach it.

Safe deploys on one host. The parts you stop doing over SSH.

Rolls back a failed release

If 3 probes in a row fail in the 30-second grace window, Caddy goes back to the previous version.

Rolled back to 67dd979now3 proxy probes failed in a row
Observing a07c3e512s30s grace window before retiring
Switched traffic14sapp-67dd979 to app-a07c3e5

Accepts only signed pushes

Every GitHub delivery needs a valid HMAC-SHA256 signature and must come from your repo and branch.

GOhandler.goGOqueue.go401 if wrong

func validSignature(secret, body []byte,
    header string) bool {
  hexSig, ok := strings.CutPrefix(
    header, "sha256=")
  mac := hmac.New(sha256.New, secret)
  mac.Write(body)
  return hmac.Equal(got, mac.Sum(nil))
}

Signs the raw body with your secret.

Gets HTTPS certificates itself

Set SLING_DOMAIN and Caddy obtains and renews a Let’s Encrypt certificate. Only /webhook is passed to Sling.

sling serve settings

4 of 5 on

HTTPS on your domainSLING_DOMAIN
Deploy on git pushSLING_WEBHOOK_SECRET
Deploy status APISLING_STATUS_TOKEN
Deploy one branchSLING_BRANCH=main
Keep deploy historySLING_DATA_DIR

Runs one deploy at a time

A file lock stops two deploys fighting over the proxy. A newer push replaces the one still waiting.

sling serve / queue

Db81e4a2push to main · waitingqueuednow
D7f30c19replaced by b81e4a2superseded8s
D9c41d0eholds the deploy locklive1m
D53a2b2fprevious, image keptstopped6m

How a commit goes live

Sling clones the commit, builds it with Docker and starts it beside the running version on a random loopback port. Three checks decide if users ever see it.

Three 200s in a row from /health, within 30 seconds
The X-Sling-Upstream header names the new container
No three failed probes in a row for 30 seconds

Read the deploy sequence

Sling beside Kamal, Coolify and Dokku. The smallest of the four, with the strictest rollback.

SlingKamalCoolify, DokployDokku

What you installOne Go binary and CaddyRuby gem and kamal-proxyDashboard, database, workersBash and nginx plugins

Checks the new version answers through the proxyPartly

Rolls back on its own during a grace window

Automatic HTTPS

Web UI and several appsOn the roadmap

Runs on Go, Docker and Caddy

Build it with make build, then run ./bin/sling with a repo and a full commit SHA. It pulls caddy:2.11.7-alpine the first time.

A Dockerfileat the repo root
A server on port 3000inside the container
GET /health200 when ready
APP_VERSIONoptional, set to the SHA

Read the full setup

Questions

What people ask before they put Sling on a server.

Ask on GitHub

Does Sling need Kubernetes?

No. It is one Go binary that drives Docker and a Caddy container on the same host.

Can I deploy a branch name instead of a SHA?

No. Sling takes only full 40-character SHAs, and checks that HEAD matches after checkout, so a moving branch can't change what ships.

How do I go back to an older commit?

Run sling rollback with its SHA. Any commit whose image is still on the host comes back without a rebuild, through the same checks as a deploy.

What if Sling is killed mid-deploy?

The next deploy marks it interrupted and records the stage it reached. Run sling status first: traffic may already be on the new version.

Can it run several apps on one server?

Not yet. Today it runs one app per host. Several apps and a web UI are on the roadmap.