ec2_snapshot question: how to run only when needed?

Archives Ansible Project
1

I'd like to snapshot EC2 volumes only when needed. I couldn't find an
easy solution that didn't require calling out to boto or other tools.
If it exists, please ignore the following :slight_smile: I wanted to ask here
before opening a feature or pull request.

It could look like this:

- local_action:
    module: ec2_snapshot
    region: "{{ region }}"
    instance_id: "{{ id }}"
    device_name: "{{ device }}"
    snapshot_tags:
        SnapshotCreator: ansible
        Host: "{{ server }}"
        Device: "{{ device }}"
    # new feature, using the snapshot_tags to determine current state
    # currently it's "always"
    state: present|absent|always

Does this make sense? Currently it's not available but could be
implemented in a backwards compatible way if the default state was
"always."

Thanks
Ted

2

Alternate suggestion for this new behavior: copy the ec2 module's count
options. It would add these options to ec2_snapshot.py:

...
  exact_count:
    version_added: "1.8"
    description:
      - An integer value which indicates how many snapshots that match the 'count_tag' parameter should be present. Snapshots are either created or terminated based on this value.
    required: false
    default: null
    aliases:
  count_tag:
    version_added: "1.8"
    description:
      - Used with 'exact_count' to determine how many snapshots based on a specific tag criteria should be present.
    required: false
    default: null
    aliases:

So when `exact_count` is omitted, things work like they do now, but if
it's specified, an `enforce_count()` function will be called.

Ted

3

...

Final suggestion :slight_smile:

boto has a nice `trim_snapshots()` method [1] that could simply be
called inline, given appropriate parameters from the ec2_snapshot
module. I could use that too, but it seems much less flexible than the
ec2 module's exact_count/count_tag approach.

Thanks!
Ted

[1] http://boto.readthedocs.org/en/latest/ref/ec2.html#boto.ec2.connection.EC2Connection.trim_snapshots

4

Hi Ted,

We’d be happy to consider a pull request to ansible/ansible-modules-core

(We would like to avoid cluttering up the tracker with feature requests as usually folks are unlikely to scratch itches of others)

ansible-devel list is the right forum to discuss what form this maybe should take.

Thanks!

5

We'd be happy to consider a pull request to ansible/ansible-modules-core

All right, as soon as I know what direction you prefer :slight_smile:

(We would like to avoid cluttering up the tracker with feature requests as
usually folks are unlikely to scratch itches of others)

That's fine, I already started developing one version, using
exact_count/count_tag. I just don't want to go too far and get the PR rejected.

ansible-devel list is the right forum to discuss what form this maybe
should take.

All right, following up there. I hope the original thread is clear from
what I quoted. My personal preference is to implement
exact_count/count_tag like the ec2 module does, but hooking into the
boto module's `trim_snapshots()` method is really effortless.

...ph, and maybe even then `trim_snapshots()` could be used through a
new ec2_snapshot_trim module. But that seems really trivial, since it's
just a call to python in the shell:

- shell:
    cmd: python -c 'import boto; import json; ec2_conn = boto.connect_ec2(); ec2_conn.trim_snapshots(8, 7, 4, True)'

Let me know...

Thanks
Ted

6

Ping?

Ted

7

I don’t really have this specific use case, so it’s hard for me to weigh in.

I’d see what others have to say that do have this particular question, and maybe it’s a new param or a new module, or maybe it’s a custom module that you use, etc.

8

I had this problem too. I didn't solve the cleaning up of old snapshots issue, but I added a feature to ec2_snapshot to only ceate a new snapshot if the last one is older than X minutes. My changes are in this PR if you're interested: https://github.com/ansible/ansible-modules-core/pull/76