Find a smoother redeployment strategy
Closed this issue · 1 comments
jamierajewski commented
Currently, the deployment strategy looks like:
- Detach important volumes:
- Bastion (for our own stuff)
- Monitor (for custom changes to the grafana dashboard)
- openLDAP (current LDAP database)
- Do NOT use terraform destroy - this doesn't seem to obey the openstack instance rule of not deleting volumes when instances are deleted
- Instead, delete the instances manually in openstack (only necessary if changes aren't detected by terraform, like a template change; otherwise, terraform apply will update the instance)
This is a bit rough and not entirely secure as one has to remember to detach the volumes for safety.
jamierajewski commented
The new strategy (for now) is to use an NFS mount to hold LDAP and Grafana/Prometheus data. This simplifies the process greatly as I can ignore these, but it takes away from the general-purpose design of the project. Will open a separate issue focused on dealing with this.