Deployment
Operations

From git push to Production: What Forte Handles for You

Follow a change from git push through its build, startup, health check, and customer test. Find the right evidence when a release fails.

Forte TeamPublished October 21, 2026Updated November 7, 20265 min read

Imagine you've pushed a new search route for a customer portal whose API runs on Forte. The build turns green, but the app fails to start a few minutes later. Now you need to find out what happened when the new image ran with the Service's production settings.

The build log can tell you whether the source was packaged successfully. When an image builds and the process fails on startup, the deployment record and startup logs are where you'll learn why.

For a Forte Service, the platform builds the selected source version, starts the new app, and records its health check. You can inspect each stage in the deployment record without maintaining a separate tool for each part of the release.

Start with a change you can find

In this example, the search change reaches Forte when you push its commit to the GitHub branch connected to the Service.

The first question is which version Forte is building. Keep the commit ID with the release so the engineer can connect the source change to the running app. If two people push changes close together, “my latest push” will not be enough to find it.

A push to the configured branch starts a deployment by default. Teams can also use the configured manual workflow. Check the Service's deployment record for the actual result. GitHub status reporting is best effort, so a stale GitHub indicator does not necessarily mean the build stopped.

Build the app with the right inputs

If the repository contains a Dockerfile, Forte uses it. Otherwise, a Smart build detects the supported app's settings. Later Standard builds reuse the detected settings so routine pushes can follow the same process.

Forte packages the app into a container image in a Linux build environment. The Service receives its runtime settings when Forte starts that image, so the app can build successfully even if a setting it needs at startup is missing.

This distinction matters for secrets. Ordinary environment variables are available as build arguments, while secrets are runtime-only. A build that tries to read a runtime secret will need a different design. Do not move the secret into source code or public build output to get past the error.

When the build fails, start with its log and the selected commit. A dependency error belongs here, before investigating the database or public domain. Builds have a documented 30-minute limit, which is another reason to find a stuck installation rather than repeatedly waiting for it.

Check how the app starts

When the portal's new process exits on startup, open its startup output and check the runtime settings for that deployment. The search change already produced an image, so a second look at the build log may tell you little about why it exited.

Check whether a required variable is missing, whether the app can reach a dependency, and whether it listens on the expected port. Read the observed error before changing several settings at once.

The app also needs an HTTP GET health route that returns a successful response. Forte discovers the port and health route during deployment. The check must be reachable without customer sign-in and should not change data.

A health route needs a clear purpose. If it checks an optional external service, that service's brief outage could prevent an otherwise usable app from deploying. If it always returns success, it may miss a dependency the app needs for every request. Decide what accepting traffic requires, then design the check around that decision.

If you change the app's port, runtime version, base image, or project layout, review the detected settings. You may need to reset the relevant setting so Forte detects it again. Routine code changes do not require that reset. Deployment guide

Find the result in the deployment record

The new app must pass its health check before Forte treats it as healthy. The record shows the ordered stages and gives the team a place to inspect the result.

Forte deployment detail for a successful release, showing a succeeded status, the source commit, and the ordered build, deploy, and health-check stages.
A successful deployment in Forte, from the source commit through the ordered build, deploy, and health-check stages.

For our example, the first search release might fail because the health route requires login. The record tells the engineer to look into health and startup behavior. After fixing the route, a new deployment should get through that stage.

Forte deployment detail for a failed release, showing a 'Service Health Check Failed' banner with guidance to inspect startup logs, a Failed status, and the build timeline stopping at the health-check stage.
A deployment that failed its health check: the failed stage is named and the failure links straight to the startup logs.

When an eligible update fails and a prior healthy revision exists, Forte's rollout can return the Service to that revision. A first deployment has no earlier version to return to. Keep those cases separate when planning a release.

An older container also cannot undo a database change. If the new code deletes a column that the old code needs, restoring the old image will not restore the column's data. Plan schema changes so the versions that may run during the rollout remain compatible.

Test the customer feature after health passes

A healthy deployment has passed its configured check. The search feature still needs a test with realistic app behavior.

Sign in with a test account, create a record, and search for it. Then try a result the account must not be allowed to read. The app should return the right records and enforce the customer's access rules.

If one of those requests fails, open the Service's request record and inspect its status, timing, and matching app logs. Include the X-Request-Id value in the app's output so overlapping calls can be matched accurately. See production request debugging.

Once the release record confirms that the process started, the customer test tells you whether the new search route returns the right results and respects the account's access rules.

Leave the next engineer a clear path

A release is easier to operate when each problem has a sensible starting point. Use the build log for packaging errors, startup logs for process failures, and request records for customer calls. Check schema compatibility when returning to an older version.

For our portal, that means the team can follow one commit through the whole release and explain why the first attempt failed. The next engineer does not have to learn a collection of private scripts before making the same investigation.

Forte supplies that common build, deploy, and diagnostic workflow. Your team supplies the app's health behavior, useful logs, feature tests, and data-change plan. Try one deliberate failure in a test setup so you know where to look before the next real one.

Take one commit through a Forte deployment

Deploy a small test app, check the result, and try a harmless failed update. Learn where to look before a real release needs help.

Deploy a test Service
How to Find the Logs for a Failed Request on Forte
Observability
AWS
Debugging

How to Find the Logs for a Failed Request on Forte

Find a failed customer request, follow its request ID into app logs, and check the fix. Learn where CloudWatch and broader tracing tools fit.

5 min read
Build Your First Paid Product on Forte
Payments
Tutorial
SaaS

Build Your First Paid Product on Forte

Build a test SaaS journey on Forte with sign-in, Postgres, payments, and logs. Check what remains before a live launch.

6 min read