Troubleshooting

Find the message you see, or the thing that goes wrong. The message of a failed deployment is on its page under Deployments. Lines from the running app are under Logs.

A quick way to find most of these before you deploy: Check your app.

Message or sign	What to do
There is no package.json at the top of the repository	Your app is in a folder. Set the Root directory in the build settings
The root directory … is not in the repository	The folder name is wrong, or it is not on this branch. Correct the Root directory
Branch … was not found	The branch was renamed or removed. Choose the branch under Settings
A command is “not found” during the build, tsc for example	The tool is not in your package.json: on your machine it is installed for every project. Add it to devDependencies. See Packages your build needs
The install fails because the lockfile does not match	Run the install on your machine, commit the lockfile, push
The install fails in a prepare or postinstall script	The script calls a tool that is not installed. Make it skip the tool when it is missing
The build took longer than 30 minutes and was stopped	Make the build shorter, or build less
The repository is larger than 1 GB	Remove large files from the repository

The deployment fails and the previous version stays online.

Message	What it means	What to do
… exited right after starting with the new image	The app crashed when it started	Read the runtime log. Often a variable is missing, or the start command is wrong
… did not start with the new image	The start command did not run	Check the Start command
… started but its health check did not pass	Nothing listened on the port in time	Listen on process.env.PORT and on 0.0.0.0, within 2 minutes
The start command could not be run	The command does not exist in the app	Check the Start command. A tool from devDependencies is not in the running app

In the runtime log you may also see:

• Waiting for the app to listen on port N. The app has started but does not listen yet.
• The app refused a connection … make sure it listens on 0.0.0.0:N, not only on 127.0.0.1. Your server listens on localhost. See Port and address.

The running app is not changed. See Deploy commands.

It was asleep and had to start. Switch on Keep one replica always on, or Wake up fast, under Scaling. See Sleep and wake.

The memory and the disk of the app are temporary. They are lost when the app is deployed or restarted, and when it sleeps. Keep data in a database or a store outside the app. See Data and files.

The app was asleep. Jobs run only while the app runs. Switch on Keep one replica always on.

Variables reach the app with the next deployment. Choose Redeploy now in the notice at the top of the app.

Line in the runtime log	What to do
The app exited with code N	Read the lines above it for the error your app printed
The app ran out of memory and was stopped	Choose a larger size under Scaling
The app kept crashing: after N restarts it is not started again	Fix the error and deploy again

The request used a hostname the app does not know.

• A domain of yours. Add it under Domains and wait until it is verified.
• The platform URL. It is switched off. Switch it on under Domains, or use your own domain.

• Compare the records at your DNS provider with the ones under Domains: type, name and value.
• Switch off a proxy or CDN in front of the record.
• DNS changes can take a while to spread. The check runs every few minutes.

• The app is paused. Choose Resume, then deploy.
• Deploy on push is off. Switch it on under Settings, or choose Deploy.
• The push went to another branch. Only the branch of the app deploys.

Write to support. The address is on the contact page. Name the app and copy the message you see, with its reference if it shows one.