10.1.0: Kubernetes, flows in Git, and premium connectors you install yourself

10.1.0 is the official release that rolls up everything shipped since 10.0.8. Three items lead it: a managed cluster on your own Kubernetes, flows in Git, and premium connectors you install yourself.

A managed cluster on your Kubernetes

Etlworks runs as a managed cluster on self-managed Kubernetes or on Amazon EKS, Azure AKS and Google GKE. Application pods are added when demand grows and drained before they are removed, so scaling in never cuts a running flow.

Day-two work happens in the app. Settings → Elastic cluster holds the scaling settings (CPU, memory and heap targets, capacity schedules, a temporary minimum), the pod size, and application upgrades, either rolling or in maintenance mode. Offline backup and restore, recovery from a failed upgrade, and migration between a regular and a managed deployment are in the installation guide. Etlworks on Kubernetes explains how the scaling works, and a 14-day Kubernetes trial is on the downloads page.

Data integration as code

Flows, connections, formats, macros and schedules can now live in Git as readable files. You review changes in pull requests and promote a project from dev to production with plan and apply, with secrets kept out of Git. It works from the app, with a Git Repository connection and deploy on push, or from CI with the new etlworks command-line client. Flows as code covers the design; the developer docs cover the commands.

Two parts are useful on their own:

• Git in Explorer. Explorer, the built-in data browser and file workspace, opens any Git repository: branches, tags and commits, file previews and diffs, and a server-side workspace where you edit, commit, push and pull. Working with Git repositories in Explorer.
• Code from Git. Any SQL or script field of a flow can read its code from a branch or tag at run time, {{git:Analytics repo:sql/orders/merge.sql@release-2026.11}}. Every file in one run comes from the same commit, including nested flows.

Run CLI Script is a new flow type that runs CLI commands as a flow: on a schedule, from a webhook, or as a workflow step. Like Run Simba Prompt, the flow type that runs our AI agent, it can only read until an administrator grants Allow changes on the flow.

Premium connectors from Settings

Premium connectors are now installed, licensed, trialed, updated and removed in Settings → Premium connectors, or from the connection that needs one. Each install is verified and reaches every server with no restart, and you can switch between versions after automatic checks. On Etlworks Cloud an administrator can install a connector and start a trial; on dedicated and self-hosted instances a super administrator manages everything. Managing premium connectors.

Connectors, CDC and SQL

• S3-compatible Object Storage is a new connector for DigitalOcean Spaces, MinIO and any other service that speaks the S3 API, with the same features as the Amazon S3 connector. S3-compatible Object Storage.
• CDC preprocessors receive the before and after row images as independent copies on every CDC connector, so a script can compare old and new values.
• Execute SQL can keep a managed high watermark ({HIGH_WATERMARK} and {HIGH_WATERMARK_UPPER}) with persistent checkpoints, and can run SELECT, UPDATE and DELETE against MongoDB.
• FTPS connections can pin TLS 1.2 or 1.3, and two opt-in upload options handle servers that fail uploads with 451 errors.

Smaller things

thread-dump is a new CLI command that prints a live thread dump of the ETL engine for diagnosing stalls. A new setting, Disable Automatic JavaScript Imports, turns off the generated Java imports for flow JavaScript. Repeated importPackage(...) calls are deduplicated, so long scripts no longer slow down in loops.

Fixes: exporting a flow built in Composer, our visual workflow builder, now includes the child flows and macros it references, and importing a connection with Replace but without credentials keeps the target's passwords.

Upgrading

One default changes. A blank Default DataSet SQL Engine setting now means Modern, which is the setting's documented default; set it to Legacy explicitly if you relied on the old behavior. Integration Agents older than 10.0 still use Legacy. The etlworks client needs a server on 10.1.0 or newer.

The full release notes link every guide.