Touchdown

-------------------

- Avoid a case where SSHing into an ASG can actually log into a random instance
in the cluster. It looks like ``DescribeInstances(InstanceIds=[])`` is
interpreted as 'any instance'.

- Make sure when bringing up an ASG for the first time it's metadata is updated
after at least one instance becomes healthy. This avoids cases where the ASG
looks empty to later steps of the deployment.

- When syncing a folder to S3 use application/octet-stream if unable to guess
mimetype.

-------------------

- We now use botocore at least 0.107.0. This version changed the structure of
its metadata folders. In order to continue overriding some values we've had
to mirror that change, which breaks support for older versions.

-------------------

- The release merges the fuselage integration refactor. You no longer have to
import bits of fuselage to use it, you can use touchdown serializers and you
can use it against your local machine. This means you can generate a
``local_settings.py`` for your Django dev environment with AWS resource
details in it!

A simple example is just::

from touchdown.core import serializers

provisioner = workspace.add_fuselage_bundle(
target=workspace.add_local(),
)

provisioner.add_file(
name="/home/john/hello",
contents=serializers.Const("HELLO!!"),
)

Thanks to mitchellrj for this feature.

- Print our resource ids after resource creation. This is particularly useful
for AMI creation and copying.

- Better SSH connection retrying when tunneling.

-------------------

- The parallel resolver introduced in 0.0.7 is now enabled by default.

- We've added more validation! Bucket names are now validated with the same
rules as the web console. You can only use storage_encrypted with a subset of
RDS instance sizes. We know capture that before making a single AWS call.

- asg: We now wait for there to be at least one healthy instance when creating
a new auto scaling group.

- asg: We now clean up old launch configs (you can only create a hundred, and
with 5 server types you can hit that in 20 deployments!).

- kms: Add initial support for managing Key and Alias resources.

- ssh: Fix streaming from paramiko channels. It was racy, and this meant that
short lived commands (such as ``whoami``) were prone to not returning all
their stdout. This mean that our connection health checks would fail.

- ssh: When using ssh-over-ssh the inner connection might fail with
ChannelException. Assuming that an instance might still be booting we now
retry in this case.

- elb: Fix failure in elb teardown.

- cloudfront: If resuming a teardown touchdown will now resume waiting for
web distributions to enter the 'deployed' state before attempting to delete
them.

- Of interested to developers: we've switched to stricter flake8 settings. We
are no longer inhibiting any of the default errors. Our ``max-line-length``
is 150 and our ``max-complexity`` is 11. We will be reducing these further in
the coming months.

------------------

- s3: Add very basic support for syncing entire folders to s3.

- rds: Add point in time restore and snashot restore.

- cloudfront: By setting cname to '' you can create a alias-less distribution.
[mitchellrj]

- Any field can now accept a serializer. This means any field can be set
dynamically based on other resources outputs.
[mitchellrj]

- We now run our tests on AppVeyor on python 2.7 (32 bit) and python 3.4
(32 bit and 64 bit).

- elb: Now waits for Elastic Network Interfaces to go away before considering
an elb deleted.

------------------

- The s3 describe step now retrieves the location a bucket was created in.

- S3 buckets are now created with a LocationConstraint based on the account
resource. For example::

aws = workspace.add_aws(region='eu-central-1')
aws.add_bucket(name='my-bucket')

Will now create the bucket in eu-central-1, like you would expect.

- Add's support for copying images created with ``add_image`` to other regions.