Troubleshooting

Some possible issues you may come across in trying to get Montagu running, especially on a local dev environment, and how to get around them.

Please add to this doc if you find and fix any similar issues.

Docker Issues

'No basic auth credentials' error when fetching image from docker

You need to make sure you're logged in to docker as the montagu user, not as yourself, and for this you'll need to have vault set up. See here for some clues: https://github.com/vimc/montagu-registry/tree/master#login

If you've been running montagu applications at all this should all be set up, but it went snafu for me somehow.

I've got a load of docker containers hanging around, how do I get rid of them all?

Run:

docker stop $(docker ps -aq)
docker system prune --force

This script is present in some repos as clear-docker.sh.

To get rid of old volumes, you can run docker volume prune --force

To remove (but not stop) containers and volumes in one line, use docker system prune --force --volumes.

A container isn't starting or seems otherwise broken. How do I see its logs?

docker logs [CONTAINER NAME]

Vagrant / VM issues

How can I run a support VM locally?

From montagu-machine, run vagrant up <machine name>

You'll probably want to use the machine name dev, whose configuration provides a minimal data set-up (similar to that used by BuildKite) in order to test code.

The other machine names you can use are uat, science and latest which are the configurations for the main support VMs. Be careful with these, as these configurations may cause sizable data transfers - you can check the settings values for each in /montagu/settings.

Run vagrant ssh <machine name> to log into the VM.

You can modify the montagu settings for deployment once you've logged in by editing ~/montagu/src/montagu-deploy.json, e.g. change initial_data_source to minimal. (The values in this file have been copied from the relevant file in /montagu/settings during VM provisioning.)

And how do I tear it down again?

If you want to destroy the machine entirely, so that it is rebuilt next time you run it, use e.g.:

vagrant destroy uat

You should also destroy the machine's disk by running /staging/scripts/destroy-disk <machine name>

If you just want to stop the machine so that it reboots as it is next time, use halt instead of destroy.

I'm trying to deploy a VM locally, and it's complaining about the api server name when setting up orderly.

The instance_name setting needs to match an api server defined in orderly's orderly_config.yml. Try changing the instance_name setting in montagu-deploy.json on the VM to 'buildkite' to bypass this step.

Other issues

Argh, it's killed my machine with a data avalanche!

Data is pulled from production if the 'update_on_deploy' flag in montagu/src/montagu-deploy.json is set to true. So you need to make sure that's false whether locally or in the VM (depending on where you're installing montagu) or it will absolutely murder your puny machine (~1TB I think).