Skip to main content

What happens after you ship

Pressing Ship closes the wizard and opens the build view. The app does not exist until this finishes.

Who can use it: the Builder role or higher.

What a build actually does​

There is no image build step in this version, which matters for reading everything below.

The platform runs your framework on an image it already has, clones your repository into the pod when the pod starts, and installs your dependencies there. So a "build" is a clone, an install and a boot. Nothing is pushed to a registry, and no artifact is produced that a later deploy reuses.

Two consequences follow. Build tips says so: dependencies install on every start, so a repository app takes at least 2 minutes to go live. And a dependency error surfaces as a build failure rather than a code error, because it happens before your code runs.

The stage strip​

Five stages are shown: Clone repo, Install dependencies, Pull runtime image (the platform image for your framework), Provision compute & secrets and Boot app & wait for health.

Only the first and last are reported by the platform. The three in the middle are inferred in your browser from what the log says. When a build fails, only the stages that actually ran are ticked, and the one that failed is marked. When the app is live, the view hands over to the app's page.

Read the strip as a rough position and the log as the truth. When they disagree, the log is right.

There is an elapsed counter rather than an estimate, because the platform does not predict how long your dependencies take to install.

The log​

The build log streams while it runs, with All, Errors and Warnings filters, and actions to copy and download it. This is the output of the clone and the install, not your app's own log; your app has not started yet.

If the stream drops, the banner reads "Lost connection to build log stream. Reconnecting…" and the log footer reads "stream interrupted". The view reconnects every few seconds by itself, without repeating lines, and a quiet build that is only installing does not drop.

After 45 seconds without a stream, the view says it lost the stream and that the build may still be running, because a lost stream is a connection problem, not a build failure. It offers View app status, Retry build and Back to App Studio. None of them cancels the build, so check View app status before starting another one.

Cancel​

Cancel build asks "Cancel this build?" first, and says the pod will be torn down and the deploy marked as cancelled. Choose Cancel build to confirm or Keep building to stay. A confirmed cancel is a real stop: the deploy appears in the history as cancelled and the view returns to the list, and you can ship again from the wizard.

When it fails​

A failed build stays on the same page with its cause, so you can read the log that produced it. It offers View raw logs, Back to App Studio, Retry build, which starts a new build of the same app and follows it, and Apply fix where a fix is suggested. Each failure category and its fix is on When a build fails.

The commonest by far is a dependency that will not install, which is a consequence of the install-at-boot model above. The second is a health check probing a path your app does not serve, which is set in Runtime and looks like a failure rather than a misconfiguration.

Describes the platform as of .

Was this page helpful?