BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
IT operations pros have some work to do to automate the infrastructure underpinning DevOps initiatives.
While cultural barriers are some of the most daunting DevOps challenges, IT operations practitioners say that capturing infrastructure as code is the most significant technical hurdle to supporting modern application development practices.
Not everyone has yet bought into the concept of taking a traditional rack and stack infrastructure with management of IPs in Excel spreadsheets, and automation through Bash scripts and Ruby code, according to Pauly Comtois, vice president of DevOps for a multi-national media company.
"A lot of our customer organizations barely have operations automated in any way," echoed Nirmal Mehta, senior lead technologist for the strategic innovation group at Booz Allen Hamilton Inc., a consulting firm based in McLean, Va., who works with government organizations to establish a DevOps culture.
"It's 2016, and we should be able to automate those deployments," he said. "Once you do that, you can start to use the exact same tools to manage the infrastructure that you use for your application code."
A big reason why companies have been slow to automate their operations is that infrastructure as code work can be more easily discussed than done -- legacy applications often weren't designed with tools such as Chef or Puppet in mind.
Third-party software that runs on Windows isn't conducive to automation via the command line, Comtois pointed out. "What makes that really technically challenging is when that piece of software also happens to be critical to the workflow of that organization, so I can't just go in and rip it out and replace it with something else."
These issues can be overcome, but "some transitions are more painful than others," he said.
Security teams also have to be brought on board with managing infrastructure as code, according to Mehta.
"Infrastructure as code and configuration management make compliance a lot easier, but that also means that compliance is no longer a thing that you do once a year," he said. "It gets enveloped in the DevOps process [just like] any piece of code needs to go through."
The majority of time spent by IT operations into the foreseeable future will be transitioning manual processes into infrastructure as code, or automated steps that follow the same pipeline that application code does, according to Mehta.
Infrastructure as code benefits
So why go through the technical headaches to establish infrastructure as code?
According to experienced DevOps practitioners, it's the only way to create an automated IT infrastructure that adequately supports automated application development testing and release cycles.
"In our environment, Jenkins makes many calls into Ansible to build stuff and deploy and configure it," said Baron Schwartz, founder and CEO of VividCortex, a database monitoring SaaS provider based in Charlottesville, Va. "Whatever we want to be automated, we have CircleCI calling a Web service that pokes Jenkins, which runs Ansible -- it sounds like a Rube Goldberg machine, but it works well."
Even things the VividCortex team wants to kick off manually use a chat bot to call into Jenkins and kick off a build job with Ansible, Schwartz said.
Getting IT ops staffs used to the concept of infrastructure as code is key to securing their buy-in as DevOps is more broadly rolled out in an environment, according to Caedman Oakley, DevOps evangelist for Ooyala Inc., a video processing service headquartered in Mountain View, Calif.
"Operations doesn't want to see things change unless [it] know[s] what controls are in place," Oakley said. "Everything being written in a Chef recipe or in cookbooks means [operations] can see what the change was and ... knows exactly who did the change and why it's happening -- and that actually is the greatest opener to adoption on the operations side."
Ultimately infrastructure as code simplifies infrastructure management, Oakley said.
"Operations can just go manage the infrastructure now, and don't have to worry about figuring out why one server is slightly different from another," he said. "You can just fire up an instance any way you want to."
What's the difference? DevOps vs. NoOps
How to nail a DevOps job interview