What’s coming up in Argo Workflows v2.12?

140 pull requests from 41 different contributors.

Reporting and Metrics

You can now create a report on historical runs of your workflow:

Resource duration is now added up for DAG and steps, so you can see how much CPU, memory, or GPU part of workflow was configured to use.

We now automatically predict the amount of time a workflow will take based on its last successful run, and show you that estimate.

We also show the current progress of your workflow. This is shown as the number of nodes in your workflow that have been completed over the total number of nodes.

SSO + RBAC

SSO has been updated to use a compressed JWE rather than a plain JWT, make SSO tokens both smaller and completely opaque to the end-sure.

We’ve also introduced fine-grained RBAC configuration options to SSO. Previously when using SSO, all users would end up using the argo-server service account. Now you can configure multiple service account and annotate them with rules that determine which to use based on the user's OAuth 2 groups.

Other Noteworthy Changes

• Workflow memorization can now have a TTL.
• Workflow archiving is now fully robust to temporary loss of connection to your database.
• When unpacking artifacts, you can now set the file mode recursively.
• You can choose when to have any volumes of garbage collected.
• You can now set headers for HTTP artifacts.
• Use started and finished in configurable links to make log facility searching faster.
• Mitigate problems with mutating webhooks by delaying the first workflow reconciliation.
• Delete workflows using a field selector.
• Several improvements for the performance of large workflows, including faster reconciliation time, and greatly reduced memory consumption.
• Improved reliability: the automatic re-creation of deleted pending pods, and the option to set a single workflow-level retry strategy,
• Improved security: we run the controller and server to run as non-root, and we’ve documented how to run workflow pods as non-root.