Ansible 1.9.0 released to take a look at this configuration management upstart _ Automatic shipping Tools

The main changes in the Ansible 1.9.0 (version code is dancing in the street,ansible each major version of a song in the Vanhallen band) are: Added Kerberos support to WinRM Connection plugin. Tags rehaul:added ' all ', ' always ', ' untagged ' and ' tagged ' special tags and normalized tag resolution. Added tag information to--list-tasks and new--list-tags option. Privilege escalation generalization, new ' become ' system and variables now'll handle existing and new methods. Sudo and Su have been kept for backwards compatibility. New methods Pbrun and pfexec in ' alpha ', planned adding ' runas ' for winrm connection. ......

More details can be consulted here.

Configuration management is also an area of choice, with more than 10 20 of them in Wikipedia, with puppet, chef and Salt and cfengine, vagrant and Nixos all familiar. But Up-and-comer ansible more popular, GitHub star incredibly more than 10,000, far higher than the puppet and chef three thousand or four thousand, salt 4953, only the indirect competition vagrant 8,000 more than the same.

In the Ansible company website, Twitter, Evernote, NASA, GoPro, Atlassian and other well-known enterprises are their users.

Why is ansible so hot? Let's start with what it's all about.

Michael DeHaan, who developed cobbler and func in Red Hat and worked at puppet in February 2012, saw opportunities in it automation: Linux administrators had to use several types of tools to cope with different jobs Configuration management is puppet or chef, deploying with fabric or Capistrano, and Func or mcollective to handle other tasks, which are generally too complex. At the same time, multiple-node deployments do not have the tools to handle well, and in the cloud and large-scale Internet infrastructure, this is the most interesting question.

One day, DeHaan started using Python to develop a new tool on his couch, and his goal was to be extremely easy to use, even if he wanted to, and anyone could learn to do it in a few minutes, and it would be easy to remember for 6 months. This is ansible.

Because DeHaan is already very famous in the Operation Dimension Circle, Ansible soon became popular after release. The adoption of Fedora's Seth Vidal (yum author, July 2013 car accident) and the publication of the April high scalability article were critical.

After this, DeHaan also participated in the development of OpenStack, but in the process of puppet automation OpenStack constantly hitting the wall. At this time, ansible in GitHub lit up. Soon he decided to set up a company--ansibleworks. The company gained 6 million investment in August 2013.

Ansible only relies on SSH, and it is extremely easy to get started without installing agents on remote machines. Hacker News Someone called (shell scripting) + +, very in place.

DeHaan later summed up the reasons for Ansible's success:

The key idea is to develop a good automated system that recognizes that computers are a group of machines that are not just separate, that is, "multi-tier choreography." The modeling process is as important as the modeling state. The idea of relying on custom agent architectures without traditional configuration management avoids certificate exchange, as well as reverse parsing of DNS and NTP issues. By default pluggable, everyone can easily contribute, thus gaining wide participation and adoption. Keep it simple (with YAML, etc.), make plans and stick to them, and then be optimistic.

Ansible 1.9.0 's version code is dancing in the street,ansible each major version of the code is a piece of Van Halen band. By the way, DeHaan has left ansible this January and is not too much involved in the ansible community, but has joined the DataStax (the company behind Cassandra) as project manager. The title of his farewell article Happy Trails is also Van Halen's song name. But he's still active on hacker news.

Today, in configuration management, ansible and salt are stars that shine. So how do they choose between them?

There is a lot of discussion on Hacker news. Two paragraphs of which illustrate the problem:

Ansible is much simpler, essentially a mixture of declarations and commands, and may be less performance on large clusters. The salt is set up much heavier, with proxies and servers, with metadata for remote machines, and many terms and concepts are their own. Ansible is more appropriate in smaller environments (20-50 machines).

Saltstack sensory function is more complete and more powerful. Ansible does not rely on graphs, sometimes it is difficult to modularize. I used it myself: ansible start salt, then let salt dry dirty live dirty.

Small Mirneville Blog has a concise introduction.