f500 / ansible-project_deploy Goto Github PK

Hi,
I don't know, how this can happen, but even though I can find the role on ansible galaxy (it shows up in the list), when I try to proceed to the role page, I'm getting "page not found".

Ansible geeft veel van de volgende warnings:
[WARNING]: when statements should not include jinja2 templating delimiters such as {{ }} or {% %}. Found: {{ X }}

After running the deploy role, it's possible to call the deploy module with state=cleanup to remove old releases. In this case it generates a new deploy.new_release_path fact.

I would expect the deploy module to stick to the first new_release it generates.

Hi,

are you open to add the accept_hostkey option to the git task?
e.g.

git: "repo={{ project_git_repo }} dest={{ project_source_path }} version={{ project_version }} accept_hostkey={{ project_git_accept_hostkey }}"

Hi there,

first of all, great role. Helps me a ton.

I would like to be able to speed up my composer install by copying the vendor dir from the old release (like: http://capifony.org/cookbook/speeding-up-deploy.html).

I was not able to put the cp command in project_pre_build_commands because this will break on the first deployment when current/vendor is not available yet.

Currently I am patching the role with the following tasks:

- name: Check if vendor dir exists
  stat: path={{ project_root }}/current/vendor
  register: check_vendor_path

- name: Copy vendor dir if exists to speed up composer
  command: /bin/cp -rp {{ project_root }}/current/vendor {{ deploy_helper.new_release_path }}
  when: check_vendor_path.stat.exists

It speeds up the deployment process quite drastically.

What do you think? Worth a PR? Any other solutions.

Cheers

For example, if I want to do a service reload and need 'sudo: yes' for that task, it doesn't get passed properly.

Just an idea, currently the cleanup part is pretty rude:

releases = os.listdir(facts['releases_path'])
releases.sort(reverse=True)
obsoletes = releases[deploy.keep_releases:]
for obsolete in obsoletes:
    changes += deploy.delete_path(os.path.join(facts['releases_path'],obsolete))

If you have keep_releases amount of failures, the next time you deploy the current release will be removed. I have two suggestions:

1. Basic: check if obsolete is symlinked by current. If so: do not remove this directory and quit with a failure message
2. Advanced: check to what directory the current is symlinked and then remove only the directories which are "older" than the linked directory. Possibly also remove all directories "later" than the linked dir, as these are probably failed builds.

I can do a little python so eventually solution 1 is feasible to implement for me. No 2 is little more complex I need some more time to implement, if I am the one who will do the work ;)

Hi Guys,

It seems latest ansible (since 2.3?) has changed the requirements for the when parameters:

[WARNING]: when statements should not include jinja2 templating delimiters
such as {{ }} or {% %}. Found: {{ project_copy_previous_composer_vendors }}

Tested on 3.2.0 of f500.project_deploy

Not sure if the change for this would be backwards compatible?

What is the reason to automatically create a system directory in the /shared folder? It isn't used in the role anymore, is it a left-over of a development version?

Hi there,

Has anyone found an easy way to do a rollback? My problem is that timestamps are different from server to server , so it's impossible to just change the -current- symlink.

Capistrano for example is creating the same timestamp on all servers.

Current example:

lrwxrwxrwx  1 root root   53 Feb  5 02:44 current -> /.../releases/20150205024358
lrwxrwxrwx  1 root root   53 Feb  5 02:44 current -> /.../releases/20150205024359
lrwxrwxrwx  1 root root   53 Feb  5 02:44 current -> /.../releases/20150205024358

Is anyone interested in that functionality?

Hi Ramon :)

Just checking again to update with composer/npm/bower: there are modules in Ansible for composer and npm. This changes the composer/npm tasks to these very simple ones:

- name: Do composer install
  composer: command=install working_dir={{ deploy.new_release_path }}
  environment: project_environment
  when: project_has_composer

- name: Do npm install
  npm: path={{ deploy.new_release_path }} production=yes
  environment: project_environment
  when: project_has_npm

These modules are maintained by the complete Ansible community, so it is easier to keep up to date with new changes in composer/npm or their respective modules. The disadvantage now is you don't have the ease of use to tune the command parameters or its binary location. For feature flags you have several options, so you have to copy/paste them over into your defaults + task yaml, but e.g. it is not possible (yet) to change the binary location of composer. Unfortunately, there is no such module for bower yet.

Change the task or not?