With Kubernetes, the native strategy for high availability is to set up multiple master replicas, and engineer a failover system that passes control to a replica when the main master fails. When the master does fail, nothing can schedule workloads. But the problem with simplicity — here and everywhere else — is that it makes things way too complex.
There are a variety of approaches to resolving this issue, and providing what the architects of these approaches politely describe as real high availability (perhaps we should call it “RHA”?). Some involve a kind of replication of the entire Kubernetes environment. Kubo is Cloud Foundry’s contribution to this emerging group of alternatives. It’s a way of leveraging the features that Cloud Foundry already uses for load balancing virtual machines, for effectively balancing traffic to multiple concurrent Kubernetes instances inside VMs.
“We’ve hit a number of issues with different approaches,” admitted Pivotal principal software engineer Konstantin Semenov, in a discussion with TNS founder Alex Williams for The New Stack Makers podcast, “but we have it running.”
To make load balancing work securely in an HA scenario, “first of all, you have to have secure communication. Meaning you have to have certificates installed,” Semenov explained. “Those certificates should advertise the same entity, so to speak. So you have to coordinate that between all your failures and reboots.”
It’s a rational explanation why you can’t just have a whole bunch of simultaneous Kubernetes masters: the containers being orchestrated need to see a single master, even when the work is being actively distributed among a multitude of masters.
Semenov is joined by Google Cloud Platform advocate Casey West, and Google software engineer Meaghan Kjelland, for a wide-ranging discussion of the roles Kubo, Cloud Foundry, and Kubernetes play in high-availability environments. Listen now to “What Kubo Brings to Cloud Foundry and Kubernetes,” a podcast episode recorded at the Cloud Foundry Summit 2017 in Santa Clara, California.
In This Edition:
1:39: What is Kubo?
7:38: Exploring why Kubo was developed, and the problems it solves when working with Kubernetes.
14:38: How Kubo and Kubernetes free developers from middleware lock-in.
17:27: Cloud deployments in Kubo.
23:37: Open Service Broker API innovations and its current status.
27:43: The technology architecture behind Kubo.
The Cloud Foundry Foundation is a sponsor of The New Stack.