apigee / henchman Goto Github PK

• Machines based on hosts list
• Subroutine to render and run each task
• use codegangsta for the cli https://github.com/codegangsta/cli

Relates to #19. Prevents testing as well.

Current it's setting them as root.

Similar to previous henchman. Number of tasks, success, etc. Maybe have counters for each host.

Getting an empty string when plan.Name is being accessed

Port it over from github.com/sudharsh/henchman

We're passing it as one of the parameters to inventoryLoader interface. This is wrong as this assumes that the transport is going to be same for all the machines. This may not be the case and they may have different configurations. Different transport among different machines is a possibility too.

Template module should allow support for pongo template.

Complex modules can be a folder that has all the dependencies as part of the module directory. The entry point of such module should be an 'exec' script. Can bundle dependencies of the module as part of its directory.

Requires changes in the executor to figure out if the module being executed is a file or a directory.

Example:

    shell: cmd="sed -i 's,conf_security_ldap.user.store.enable=true,conf_security_ldap.user.store.enable=false,g' {{ vars.edge_mgmt_token_file }}"

Error:

2015/10/16 17:52:50 Module shell: Module args are invalid

Due to this issue,
golang/go#6376

Look into both the ansible and salt way. Hammock drive develop this!

Come up with a build script that produces tarballs for distribution for multiple archs. It would have the modules directory as part of the tarball as well and should be self-contained.

Setting a custom modules part gives a module not resolved error all the time

• Also make tests for invalid cases
• Check test coverage to see test cases we may have missed
• Check out table driven tests

Have a .retry file of some sort

Use them as enumerators instead.

1. Allow variables to be defined at group level
2. Allow individual host level overrides
3. Allow overrides of 'henchman' specific variables starting with 'henchman_< > ' ( henchman_user for example)

❯❯❯ go get -u github.com/apigee/henchman
warning: code.google.com is shutting down; import path code.google.com/p/go-uuid/uuid will stop working
warning: code.google.com is shutting down; import path code.google.com/p/go.crypto/ssh will stop working

code.google.com/p/go-uuid/uuid => https://github.com/pborman/uuid
code.google.com/p/go.crypto/ssh => https://godoc.org/golang.org/x/crypto/ssh

Currently task run does an 'exec', 'put', 'exec'. This sequence is not valid for all kinds of modules. We need to define a way of doing it
Eg. modules like 'copy' that copy stuff from one location to another on the box or by default copy from the henchman machine to the remote machine just do a put and exec ( when permissions are changed)

Where the key is the group name that can be globbed and the value is the list of all machines

Currently the YAML unmarshaller (including the custom ones) cannot handle multiple include cases:

name: "Sample plan"
vars:
  hello: world
  foo: scar
  include:
    - test/vars/vars.yaml
  include:
    - test/vars/vars2.yaml
  spam: eggs
tasks:
  - name: Sample task that does nothing
    action: cmd="ls"
  - name: Second task
    action: cmd="echo"
    ignore_errors: true

This won't throw a "multiple includes error" it will just take any value from any include field (depends on which value was added last) as long as the include value follows the right format.

Requires completion of main.go

Inventories should follow a...

---
groups (required):
  group_name (required):
    hosts (required):
      - "192.168.1.1"
      - "192.168.1.2"
    vars (optional):
      ulimit: 300
      henchman_keyfile: "~/.ssh/ssh_key"
  app:
    hosts:
      - "1.1.1.1"
      - "1.1.1.4"
hostvars (optional):
  "192.168.1.1":
     keyfile: "~/.ssh/another_key"
     henchman_user: "root"
     ulimit: 240
     files: 240
  "1.1.1.1":
     ulimit: 240
     files: 240

Using only std python libs (httplib, urllib, urllib2).
httplib - is a parsing nightmare
urllib/urllib2 can only handle GET and POST

OR

Find a way to use without stdlibs that won't cause dependency probs.
Checkout issue #44

For now have it use basic RESTful calls.

result["output"] = {'status': status_code, 'headers': headers, 'body': body}

Put it up in the wiki

Be able to specify groups, homedir..etc

Setting a custom modules part gives a module not resolved error all the time

Needed for local_action

It's being unused currently