Kubernetes: Argo Workflows 3.0 relies on a new user interface and hot standby

The Argo project has presented the third major version. Argo Workflows is a declarative tool for GitOps and Continuous Delivery (CD) that is intended to consolidate the management of configurations and definitions of the hosted applications in Kubernetes environments. Intuit and Red Hat originally launched the project in 2017, but a year ago they handed it over to the Cloud Native Computive Foundation (CNCF) for incubation and further development.

The user interface of Argo Workflows 3.0 contains new React components under the hood, which should make the interaction with the work processes easier to handle and, according to the editors, make the UI more stable and reliable. New features have been added, such as an event flow page that illustrates the integration of the workflows with Argo Events, the Argo developers have revised the log view of the workflows and, optionally, users can now embed widgets with which, for example, they can check the progress of the individual workflows – Steps or the display status can be checked. The APIs for Argo Events are new, and the release supports Go modules.

According to the CNCF’s blog announcement, the real highlight is the high availability of the controllers. In previous versions, the controller had to restart Kubernetes the pod every time after errors, which led to annoying downtimes. To get the problem under control, Argo Workflows 3.0 offers a hot standby: A second pod runs parallel to the main pod. If the first pod fails, the second takes over automatically, which means that there is less downtime. Since the term “pod” can have several meanings in IT, here is an explanation: In the context of Kubernetes, a pod represents the individual instance of a running process in the cluster. The pods referred to here are the smallest objects that can be provided in Kubernetes.

Argo Workflows 3.0, screenshot of the new event flow page

Event Flow is a proxy for events that is based on the pod logs of the event sources. The new major version of Argo Workflows contains a visualization of the event flows.

(Image: Argo)

With version 3.0, Argo Workflows also receives a standard reference for the repository, and the so-called “key-only” artifacts are new. Users should be able to adapt the configuration of the artifact repository from the workflow specification. This apparently means that standard settings can be configured across the board for an entire namespace, instead of having to make the definition for each workflow individually. The key can be used to define artifact patterns that, according to the announcement, purify the workflows through reduced mapping. This means that parameterized inputs and outputs can be omitted, and according to the blog entry, less storage space is required.

The release includes a number of other changes, which can be found in detail in the release notes and accompanying blog announcements. More information about the current release can be found see Argo’s blog entry. Also the Cloud Native Computing Foundation (CNCF) has presented the new main version on their website. The full changes can be found listed in the release area at GitHub. At the same time as the release announcement, the Argo team also announced that version 2.12 would be included in long-term support. Although the LTS version should no longer receive any new features, it should still receive bug fixes for at least six months.

Argo Workflows is a cloud-native engine that can run tens of thousands of concurrent processes, each with thousands of workflow steps. The workflow engine can be used for batch processing, serverless computing and a number of CI / CD (continuous integration / continuous delivery) processes. Institutions such as the Geneva-based European Organization for Nuclear Research CERN use Argo for 3D rendering in the CoreWeave program (on a cluster with 1000 nodes and 6000 graphics processors), and the GitOps / CD tool is also used in the area of ​​machine learning and data processing. Argo has been written in the Go programming language since 2018, and the project now includes around 400 contributors, several thousand commits, over a thousand forks and, according to the editors, has consistently high ratings on GitHub. The community channel on Slack is followed by over 5000 members.


To home page