Release 2026-05-01

This release introduces node maintenance workflows, stronger VM input validation, and a broad round of reliability and security improvements across the platform.

What's New

Node Maintenance

You can now safely take a node out of service directly from Fluid. Open a node and put it into maintenance to cordon it (stop new workloads scheduling onto it) and drain it (move running workloads off), all without leaving the UI.

Live drain status --- watch a drain progress in real time and see exactly which pods are still being moved.
Remaining-pod visibility --- the drain view lists the pods that are yet to be evicted, with the reason each one is still there, so you always know why a drain is waiting.
Built-in guard rails --- Fluid won't let you cordon the last healthy node in a cluster, protecting you from accidentally taking your workloads offline.

Safe by design

Drains run with resilient eviction and serialized migrations, so concurrent maintenance actions won't fight each other. If a drain can't complete, the status now reports the reason instead of failing silently.

VM Improvements

Resource input validation --- creating or modifying a VM now validates CPU and memory inputs up front, so invalid or non-positive values are caught before they reach the cluster.
Unique VM Group names --- VM Group display names must now be unique, and you'll get a clear error when a name is already taken instead of ending up with duplicates.

Other Changes

A broad round of reliability, security, and quality improvements:

Clearer error messages --- a reworked error-handling layer gives more consistent, human-readable messages across the app and removes duplicate or misleading notifications.
Off-site backup foundations --- groundwork for off-site ETCD snapshot backups landed behind the scenes.