Deploying
Logs, restarts and health
See what your app is doing, start and stop it, and tell OceanLaunch how to know it's ready.
Logs
- Build logs: every deployment keeps its own log. Open a deployment to follow it live, or read it later.
- App logs: the app's Logs tab shows what it prints, live: the newest 100, 300 or 1,000 lines. You can download them.
- Log history: switch the Logs tab to History to search the last 14 days, across deploys and restarts, by time range, level and text. Download saves every line of the chosen range that matches, oldest first, as a text file of up to 200,000 lines; for more, choose a shorter range.
CPU and memory
Monitoring in the sidebar shows every app's and database's CPU and memory over the last hour, 24 hours or 7 days, sampled every minute. An app's Metrics tab shows just that app. CPU is shown as a share of its own limit (100% means it's using all it may), and memory against its limit, so you can see when an app needs more or has a leak. Hover a chart for exact values, or open Show as a table.
Usage alerts
On the Pro and Team plans, an app's Settings has Usage alerts. Choose a CPU level, a memory level or both, and over how many minutes (5 to 60). When the app's average over that time is at or above a level, the workspace's owners and admins get an email, and one more when it's back under. CPU is counted as on Monitoring: 100% is the app's own share, and it can go above that for a while. Owners and admins change the levels.
Start, stop and restart
- Stop stops the app. Its address shows an error until it's started again.
- Start and Restart deploy the version that was running again: the same upload, or the same commit. Like every deploy, this has no downtime.
- Redeploy deploys the newest version: your latest upload, or the branch's newest commit.
These wait while a deployment is running. A newer deploy replaces one that's still waiting in the queue.
Console
The app's Console tab runs a command inside the running app, as the app's own user: list files, check a setting, run a database migration or a one-off script. Type it and choose Run; the output and exit code show below. The arrow keys bring back earlier commands.
- Each command gets 30 seconds and shows up to 256 KB of output. There's no terminal and no input, so interactive
programs (editors, prompts,
top) won't work. - Commands run with
sh, so the app's image needs a shell; very small images may have none. - Files written outside the app's storage are lost on the next deploy.
- Only workspace owners and admins have the console, because a command can read the app's secrets. Each command is recorded in the security log, without its text. Static sites have no console.
Files in storage
An app with storage (Settings → Storage) has Browse files: its storage folder as the app sees it. Open folders, download a file, upload files (several at once), make folders and delete files or empty folders. Uploads and downloads go through the running app, so it must be running, and each file can be up to 50 MB. Changes are made as the app's own user, and only inside the storage folder. Owners and admins can use it; every upload, download and delete is recorded in the security log.
Background workers
A background worker is an app with no web page: a queue consumer, a scheduler, a bot. Choose Background worker under Kind when you create the app. It can't be changed later; create a new app for the other kind.
- What it runs: its start command, like
node worker.js,python worker.pyorphp artisan queue:work. PHP workers need one. Python workers without one use theworker:line of aProcfile; Node.js workers usenpm startor the main file, as web apps do. JAR, Dockerfile and Docker image apps run as they would as web apps. - No address: it gets no web address, so there are no domains, password protection, IP access or uptime checks, and it needn't listen on a port. It's ready as soon as it starts, with no health check.
- Stopping: on a redeploy, stop or restart it gets
SIGTERM. Finish the current job and exit promptly; one that's still running after a short grace period is killed. Commands joined with&&,;or|get the signal too. - Everything else works as for any app: logs, metrics, console, variables, storage, databases and deploy on push.
Static sites and WAR apps can't be workers. A worker and a web app that share a queue are two apps: give both the same database or Redis variables.
Port and health check
Your app must answer HTTP on its port. That's 8080 unless you change it under Settings → Runtime (between
1024 and 65535; for JAR and Git apps only, since Tomcat and static sites always use 8080). OceanLaunch sets
PORT and SERVER_PORT to it.
A new version only replaces the old one once it's healthy:
- By default, healthy means it answers HTTP on its port with anything, even a 404.
- Set a health check path under Settings → Runtime (for example
/actuator/health) and it must answer that with 200.
It's checked every 10 seconds, starting 10 seconds after it starts, for up to 6 minutes; it's ready as soon as one check passes. If it never becomes healthy, the deploy fails and the old version keeps serving. (Apps created before 4 November 2026 keep their earlier window: a minute to start, then up to 2 more minutes of checks.)
Once it's running, an app counts as unhealthy after 6 minutes of failed checks. Uptime alerts notice an outage sooner, within about 3 minutes.
Deploy notifications
On the app's Settings tab, owners and admins choose how to hear about deploys:
- Email owners and admins when a deploy fails (on unless you untick it): with the reason and a link to the build log. The version that was running keeps running.
- A webhook told of every deploy that goes live or fails. Paste a Slack or Discord incoming-webhook address
and each deploy arrives as a message; any other HTTPS address receives JSON with
event(deploy.succeededordeploy.failed),app,workspaceanddeployment(number, commit, reason, page). Send a test checks it. - Each request carries
X-OceanLaunch-Signature: sha256=…, the HMAC-SHA256 of the body with the secret shown under the address, so your service can check it came from OceanLaunch. - Webhook addresses must be
https://on the standard port and lead to the public internet.
Password protection
To keep a staging site or a private tool to yourselves, open the app's Settings tab and give a username and password under Password protection (owners and admins). After the next deploy, every visitor is asked for them first (HTTP basic auth, at the proxy, before the request reaches your app), on all of the app's addresses. The password is kept encrypted and not shown again; set a new one to change it, or choose Remove protection and redeploy.
IP access
Under Settings → IP access, owners and admins can limit who reaches the app by IP address, one address or
range per line (IPv4 or IPv6, like 203.0.113.7 or 203.0.113.0/24):
- Only allow: when it has entries, only those addresses can open the app; everyone else gets
403 Forbidden. Leave it empty to let everyone in. Up to 50 entries. - Block: these addresses always get
403 Forbidden, even inside an allowed range. Up to 100.
It's applied at the proxy, on all the app's addresses, within half a minute, with no redeploy, and works together with password protection. It doesn't affect this dashboard, so an allow list that leaves out your own address can always be fixed here. Your other addresses still redirect to the primary domain, where the rules apply.
Uptime and alerts
While an app runs, OceanLaunch requests its address every minute, the way a visitor would: its health check path if
it has one, otherwise /. Any answer below 500 counts as up (a redirect or a login page too); a 5xx
error, no answer within 10 seconds or no connection counts as down.
- After 3 failed checks in a row (about 3 minutes), the workspace's owners and admins get an email that the app is down, with the last answer it gave. When it answers again, they get one saying it's back and how long it was down.
- The Uptime box on the app's page shows whether it's up now, its uptime over the last 24 hours, 7 days and 30 days, and its recent outages.
- Owners and admins can stop the emails for an app with Stop emails when it's down in that box.
Checks pause while an app is stopped, and while a static site is in
maintenance mode. Only the app's own
*.oceanlaunch.app address is checked, not its custom domains.
When an app crashes
OceanLaunch checks every app every 30 seconds. If it keeps restarting (3 times within 15 minutes) or has stopped by itself, it's marked Crashing, and the app page shows the most likely reason from its log, such as Spring Boot's "APPLICATION FAILED TO START" message.
Common causes:
- a missing variable, such as a database URL (see Variables and secrets);
- the app listening on another port than the one in Settings;
- not enough memory: Java stops when it runs out, and the log says
OutOfMemoryError. Files in/tmpcount as memory too (see Disk).
The sandbox
Every app runs in its own gVisor sandbox: a kernel of its own between the app and the server, the way Google Cloud Run runs containers. Your app can't see other apps or the server, and almost nothing it does reaches the server's own kernel. Apps don't need changes for it. A few things work differently than on a plain Linux machine:
/tmpis in memory (see above);- programs that load kernel modules, use raw sockets or change system settings don't work, and apps get no Linux capabilities either way;
- apps that make very many small system calls (heavy file or network churn) run somewhat slower.