Noticing problems with node_modules/
or package versions? Try
running this command to clear the local package cache.
rm -rf node_modules ./client/*/node_modules
pnpm install --force
Validating package.json...
error @: The engine "node" is incompatible with this module. Expected version "^v14.7.0". Got "14.5.0"
If you see an error like this you need to upgrade the version of node installed. You can do this with nvm use
and then following the prompts to install or update to the correct Node.js version.
This means the frontend
server failed to start, for some reason. Look through the previous logs for possible explanations, such as failure to contact the redis
server, or database migrations failing.
If you see this error when opening the app:
500 Internal Server Error template: app.html:21:70: executing "app.html" at <version "styles/styl...>: error calling version: open client/web/dist/main-ABCD.css: no such file or directory
that means the web builder hasn't finished compiling the styles yet (it takes about 3 minutes).
sg start
may ask you to run ulimit to increase the maximum number
of available file descriptors for a process. You can make this setting
permanent for every shell session by adding the following line to your
.*rc
file (usually .bashrc
or .zshrc
):
# increase max number of file descriptors for running a sourcegraph instance.
ulimit -n 10000
On Linux, it may also be necessary to increase sysctl -n fs.inotify.max_user_watches
, which can be done by running one of the following:
echo 524288 | sudo tee -a /proc/sys/fs/inotify/max_user_watches
# If the above doesn't work, you can also try this:
sudo sysctl fs.inotify.max_user_watches=524288
If you ever need to wipe your local database and Redis, run the following command.
./dev/drop-entire-local-database-and-redis.sh
We use Caddy 2 to setup HTTPS for local development. It creates self-signed certificates and uses that to serve the local Sourcegraph instance. If your browser complains about the certificate, check the following:
-
The first time that Caddy 2 reverse-proxies your Sourcegraph instance, it needs to add its certificate authority to your local certificate store. This may require elevated permissions on your machine. If you haven't done so already, try running
./dev/caddy.sh reverse-proxy --to localhost:3080
and enter your password if prompted. You may also need to run that command as theroot
user. -
If you have completed the previous step and your browser still complains about the certificate, try restarting your browser or your local machine.
When running Caddy on WSL, you need to manually add the Caddy root certificate to the Windows certificate store using certutil.exe.
# Run inside WSL
certutil.exe -addstore -user Root "$(find /usr/local/share/ca-certificates/ -name '*Caddy*')"
This command will add the certificate to the Trusted Root Certification Authorities
for your Windows user.
Firefox on Windows and macOS will not look for enterprise roots by default. If you are getting certificate errors in Firefox only, toggling security.enterprise_roots.enabled
on in about:config
should fix the problem.
If you see a certificate expiry warning you may need to delete your certificate and restart your server.
On macOS, the certificate can be removed from here: ~/Library/Application\ Support/Caddy/certificates/local/sourcegraph.test
On first install, the program will use quite a bit of bandwidth to concurrently download all the Go and Node packages. After packages have been installed, the Javascript assets will be compiled into a single Javascript file, which can take up to 5 minutes, and can be heavy on the CPU at times.
After the initial install/compile is complete, the Docker for Mac binary uses about 1.5GB of RAM. The numerous different Go binaries don't use that much RAM or CPU each, about 5MB of RAM each.
If you're running macOS 10.15.x (Catalina) reinstalling the Xcode Command Line Tools may be necessary as follows:
- Uninstall the Command Line Tools with
rm -rf /Library/Developer/CommandLineTools
- Reinstall it with
xcode-select --install
- Go to
sourcegraph/sourcegraph
’s root directory and runrm -rf node_modules
- Re-start the dev environment (
sg start
)
The Grafana and Prometheus containers need group read access for specific files. Otherwise, you will see errors such as:
grafana | standard_init_linux.go:228: exec user process caused: permission denied
prometheus | standard_init_linux.go:228: exec user process caused: permission denied
or
prometheus | t=2021-05-26T10:05:26+0000 lvl=eror msg="command [/prometheus.sh --web.listen-address=0.0.0.0:9092] exited: fork/exec /prometheus.sh: permission denied" cmd=prom-wrapper
prometheus | t=2021-05-26T10:05:26+0000 lvl=eror msg="command [/alertmanager.sh --config.file=/sg_config_prometheus/alertmanag
If files do not normally have group permissions in your environment (e.g. if you set umask 077
), then you need to:
-
Set group read permissions for the Grafana and Prometheus docker images, with
chmod -R g=rX docker-images/{grafana,prometheus}
-
Run
sg start monitoring
so that new files are group readable. If you haveumask
set to a different value you can change that value just for this command by starting it in a subshell:(umask 027; sg start monitoring)
When trying to install sg
with the pre-built binaries on WSL2 you may run into this error message: failed to set max open files: invalid argument
. The default configuration of WSL2 does not allow the user to modify the number of open files by default which sg
requires to start. To work around this you can modify the file limits for your given session with sudo prlimit --nofile=20000 --pid $$; ulimit -n 20000
then re-run the installation script.
Note: this change will be reverted when your session ends. You will need to reset these limits every time you open a new session and want to use sg
.