Use tee in GitHub Actions to save logs from a process

Here is the situation: a step in GitHub actions runs a process, in this case a Helm command, and I need to redirect the output.

The obvious way to do it is to redirect into a file

helm ... > helm.log
# show logs after done
cat helm.log

However, what happens is that GitHub Actions immediately skips this step. It doesn't wait for the redirected command to end! I also lose the ability to watch logs, especially if --wait is passed to Helm.

Introducing `tee`

tee is a command that redirects output and duplicates it. This way the output is both standard output and, in this case, a file.

Here is the new command:

helm ... 2>&1 | tee helm.log

Sweet, now GitHub Actions will wait for the process as we expected. We also get live logs.

However, if helm returns with a nonzero value, the step is a success. Which we clearly don't want.

Use `pipefail` to catch pipeline errors

Since | is a pipe, it kinda swallows errors. We need to propagate them "out".

# ensure that a step failing will make the pipeline appear to fail
set -o pipefail

# tee to split into file
helm ... 2>&1 | tee helm.log

# do something with helm.log

# restore the previous state
set +o pipefail

# Now fail if helm command returned a nonzero exit code
if [ $? -ne 0 ]; then
echo "Helm deployment failed"
exit 1
fi

Voilà! We have a functional step that will properly redirect output and fail.

Introducing tee

Use pipefail to catch pipeline errors

Introducing `tee`

Use `pipefail` to catch pipeline errors