| Commit message (Collapse) | Author | Files | Lines |
|---|
|
This way I can dynamically control whether to destroy and recreate all the
existing infrastructure entirely from scratch.
The advantages of doing so are:
- test the non-existence of local state on every deployment;
- make sure I can always recreate everything from scratch.
The disadvantages are:
- slower deployment times;
- longer downtime during deployments.
|
|
|
|
The deployment is not quite working, and I'm unable to test right now:
DigitalOcean is returning 503 for my requests.
As of this commit, I can run =ansible-playbook provider.yml= more than once and
it will actually be idempotent.
Notes:
- SSH fingerprint are now taken from the public key file instead of manually
supplying it in the terraform template using the =digitalocean_ssh_key=
resource;
- use Ansible instead of ad-hoc Bash scripts for provisioning the Droplets
created by Terraform;
- use the =filename.env.extension= to create the concrete files in CI;
- use the =user_data= to add the know SSH key pair to the newly created Droplet;
- add =rotate-ssh-keys.sh= utils;
|
|
|
|
|
|
|
|
|
|
Create a new backup entry before running =terraform apply=, which may (or may
not) destroy the current machine.
This shouldn't be an issue for the backup itself, since all of the data should
be stored in a separate Block Storage Volume, but we can take advantage of the
sevices already needing to be taken down in order to perform a full backup of
the data.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
