Kubecon 2019 If one were looking for an indicator that Kubernetes is maturing (other than the sheer numbers that turned out for Kubecon San Diego), it is the plethora of backup solutions emerging for the orchestration technology.
Attendance for the once-an-upstart tech's shindig hit 12,000 this year, overwhelming the San Diego Convention Center's facilities at times (notably after the Day One morning keynote). The December 2018 event had 8,000 eager beavers turn up in Seattle and, of course, the first conference in San Francisco back in 2015 had a mere 550 early adopters show up.
The Cloud Native Computing Foundation (CNCF) has also passed the 500-member mark with the likes of Arm signing up for Platinum status.
OpenStack backerupper Trilio and veteran (at least as far as the short life of Kubernetes is concerned) storage specialist Portworx were on hand at this week's Kubecon to punt previews of their take on keeping those precious pods backed up.
Trilio, which already has form in the OpenStack arena, has turned its attention to Kubernetes with TrilioVault for Red Hat OpenShift, replete with Red Hat certification. CEO David Safaii told The Register the company wants to make Trilio "part of the definition of the application" via a single line in the associated Helm Chart.
"That's all it takes to deploy Trilio," he said.
The core Trilio platform, with its agentless technology and incremental backup schemas, is certainly an interesting option, and something with which the Open Stack crowd will already be familiar. The Trilio approach also works for across physical infrastructure, be it on premises, in the cloud or a combination.
And the backup files themselves? Stored in the open QCOW2 format, thus eliminating vendor lock-in.
As far as cost is concerned, the Trilio gang told us that the freemium model is the way forward, with the DevOps side constrained by time and scale.
While it's Red Hat only for the time being (and very much a preview), Safaii assured us other platforms will follow in the early part of 2020.
Also lobbing a Kubernetes backup product into preview is San Francisco-based Portworx. The outfit, founded by the gang behind Ocarina Networks, is better known for running stateful containers in production.
PX-Backup is all about grabbing all the application data, configuration and Kubernetes objects into a single unit and shoving it into S3-compatible storage. The tech will also handle multi-node distributed databases such as MongoDB as well as stashing extra metadata to keep compliance officers happy.
The approach differs from the Trilio's take on things, as CTO Goutham Rao said: "It's multiple containers that are running on different machines. So when you come in and say, 'backup my Cassandra,' what does it mean?
"So I have to coordinate point in time, the consistency of this application stack across these three machines. That's not possible to do it without the data platform that understands that."
"And Trilio," Rao sniffed, "doesn't have that."
Portworx and Trilio are not the only backup games in town. The likes of LA-based Kasten will also cheerfully take money from customers in return for enterprise peace of mind when it comes to backup, restore, disaster recovery and moving of Kubernetes applications.
A bevvy of Autopilots
Recognising its storage roots, Portworx also dropped a GA version of its capacity management tool PX-Autopilot, not to be confused with solo.io's open-sourcing of its Service Mesh Operators Framework, Autopilot.
If only there was some platform that could, er, orchestrate how companies in the Kubernetes space name their stuff.
PX-Autopilot is, Rao told us, all about optimisation by avoiding the over-provisioning of block storage. Despite the promise that customers would only pay for what they use in the Kubernetes world, Portworx reckons that many overpay because provisioning additional storage can be a serious pain.
PX-Autopilot, another add-on for Portworx Enterprise 2.3, aims to address this by allowing users to set up rules for automatic scaling when storage is running low. As well saving the costs inherent in over-provisioning, the company claims that hardworking admins will have their workloads reduced as Portworx's finest does its stuff.
More interestingly, the gang has a further update waiting in the wings to shrink things down as well. Rao said: "Autopilot can kick in and say: 'That's not been used for 15 days, I'm going to tier it colder storage, maybe object storage'."
While the former PX-Autopilot functionality has gone to general availability, the latter is expected in technical preview form in the next few weeks.
Making Mesh adaptive
And the other Autopilot? Fresh from unleashing Gloo Enterprise 1.0 the previous week – with LDAP support and a built-in Web Application Firewall for, er, gluing enterprise infrastructure together – solo.io's CEO, Idit Levine, took to the Kubecon stage to introduce the open-sourced Autopilot.
This one's aimed at turning a service mesh into an adaptive service mesh, able to adjust its configuration based on changes in the environment.
To her credit, Levine insisted that the adaptiveness was rules-based, laughing that "no one here will take me seriously" if solo.io followed the depressingly common industry norm of slapping the "AI" or "machine learning" tags on the tech.
Well, quite. ®