How to work with container backups

Containers require different backup procedures than traditional servers or even virtual machines, but it's going to take a while for the new ways to be sorted out.

olm26250istock-516764040.jpg

Image: Olivier Le Moal, Getty Images/iStockphoto

The process for backing up containers is different from backing up virtual machines or conventional servers, and there's a lot to learn if your perspective has always been regular storage, experts say.

Traditional vendors that make backup software--think of IBM, EMC, Veritas, Commvault, and so on--are starting to figure out their container backup strategies, while startups such as Kasten are attempting to specialize in it.

It's vital to understand the differences. Two stand out--the people using the software, and the contents of which data is being protected.

"I'd start with the 'Who are the users?' question," said Greg Van Hise, storage software technical strategist at IBM. "It's the application developers, it's the Kubernetes administrators who we talk about here. They're interested in operating in the same kind of environment that they're used to," unlike traditional backup processes, which would be controlled by IT staff.

SEE: Implementing DevOps: A guide for IT pros (free PDF) (TechRepublic)

On the content side, "You need to focus on the complete application. You can't just focus on the disks or the volumes underneath it," explained Niraj Tolia, CEO of Kasten. Regular backup software that's made for servers or even for virtual machines can't handle applications live across clouds, he noted. Also, where most backup programs operate by putting agent code on every server, that's not a scalable approach to work with containers that tend to be rolled out in larger quantities, he said.

Another consideration is the Container Storage Interface (CSI), which debuted two years ago. The idea behind CSI is that all mainstream brands of orchestration engine, which arranges resources between containers and hardware, could communicate with storage in the same way. Standards that derive directly from the storage industry tend not to succeed, but CSI comes from the container companies themselves and seems to be gaining traction, IBM's Van Hise stated.

IBM, for its part, has a small amount of off-the-shelf container backup ability in its Spectrum line of backup software--currently the company uses VMware support to capture volumes associated with Kubernetes clusters, and they can do some scripting for coordination. Later this year IBM is releasing new software that is more Kubernetes-specific. For now, van Hise said he wants customers to get educated about use cases, recovery times, and recovery points.

SEE: How secure are your containerized apps? (ZDNet)

Both Van Hise and Tolia said container backup data can be put to use while it's waiting around to be needed in the usual way. It can be shared as a developer's sandbox for testing new code so the original is not disturbed. That trend is also happening with traditional backups, but perhaps not as quickly.

Kasten (its name is German for container) isn't the only startup working on this--another one is Nuvoloso (from the Italian for cloudy), which remains stealthy until later this year. Nuvoloso CTO Stephen Manley said he thinks containers will need more time to evolve from being new-wave development tools to becoming standard IT practice. Networking and security are their own challenges in addition to backup, he noted. "It'll take years," he said. "There's so many bowling pins to knock down on the way there."

Also see