The release of the free PaaS platform Cozystack 0.22.0, built on Kubernetes, is available. The project aims to provide a ready-made platform for hosting providers and a framework for building private and public clouds. The platform is installed directly on servers and covers all aspects of preparing the infrastructure for providing managed services. Cozystack allows you to launch and provide Kubernetes clusters, databases, and virtual machines. The platform code is available on GitHub and is distributed under the Apache-2.0 license.
Talos is used as the underlying technology stack. Linux and Flux CD. Images with the system, kernel, and necessary modules are pre-built and updated atomically, eliminating the need for components like dkms and a package manager, and ensuring stable operation. A simple installation method is provided in an empty data center using PXE and debian-similar installer talos-bootstrap.
The platform includes a free implementation network infrastructure (fabric) based on Kube-OVN, and uses Cilium for service mesh organization and MetalLB for service advertising. Storage is implemented on LINSTOR, which uses ZFS as the underlying storage layer and DRBD for replication. A pre-configured monitoring stack based on VictoriaMetrics and Grafana is included. To launch virtual machines KubeVirt technology is used, which allows you to run classic virtual machines directly in Kubernetes containers and already has all the necessary integrations with the Cluster API for launching managed Kubernetes clusters inside a hardware Kubernetes cluster.
The new version adds cozystack-controller and new entities: Workload and WorkloadMonitor — which allow you to monitor the state of pods managed by operators and evaluate the service level according to predefined rules. Applications in Cozystack are managed by different operators, so it was decided to create a single format for displaying the status of each service.
How it works: When deploying an application, WorkloadMonitor is also installed, which monitors the state of pods by selector. As soon as one of the pods falls under the selector, a new Workload entity is created for it, which displays the role for each pod and its state. In the WorkloadMonitor status, you can see the number of existing replicas and the minimum number of replicas required to service the service. As soon as the workload becomes less than the minReplicas value for WorkloadMonitor, the service is marked as non-operational.
For applications that cannot specify an exact minReplicas value (for example, Kubernetes workers can scale automatically), it is now possible to not specify this value in WorkloadMonitor at all. In this case, WorkloadMonitor will simply count the total number of running instances.
This mechanism allows you to use any operators and pod management mechanisms in Kubernetes and easily extend the platform by providing a single interface for displaying the current state of the service.
WorkloadMonitor for collecting information about replicas and their health has been added to Kubernetes, Postgres, Monitoring, VirtualMachine, VMInstance, Redis, etcd, and SeaweedFS applications. Cozystack Dashboard now displays the number of application replicas and the service level for each Workload group.
Other changes:
- The telemetry client and server have been implemented and published under the Apache 2.0 license. Telemetry metrics are collected in accordance with the recommendations of the "LF Telemetry Data Collection and Usage Policy" and can be disabled using the "telemetry-enabled: false" option in the Cozystack configuration. The data itself is anonymous and completely depersonalized. A public Dashboard based on telemetry data is planned for the next releases of the platform.
- The cluster-autoscaller component for Kubernetes has been updated, as well as its settings, which allow for more efficient scaling of clusters, both up and down.
- The MAINTAINERS file has been updated to list the project maintainers and their areas of responsibility.
- Cozystack has added a builder service application that allows you to build the platform directly in Kubernetes.
- For VictoriaMetrics DBMS, the ability to specify custom parameters has been added and the default limits have been increased.
- Collection of metrics from databases has been implemented for Grafana and Alerta.
- Added alerts about the status of virtual machines and postgres clusters.
- Metrics collection has been configured for KubeVirt and Grafana dashboard has been added.
- The extra-keycloak-redirect-uri-for-dashboard option has been added to the Cozystack configuration, allowing you to configure additional redirects for Keycloak.
- Fixed a bug in VMInstance that blocked VMdisks from connecting to virtual machines.
- Updated Flux Operator 0.12.0, Flux Instance chart 0.12.0, Cilium 1.16.5, Kube-OVN 1.13.2, CNPG PostgreSQL Operator 1.25.0, Talos Linux 1.9.1.
Source: opennet.ru
