Upgrading to 1.28.0
Released 2026-11-02. Everything in the release: changelog. How to upgrade: Upgrade.
What's new
- A Kubernetes cluster is managed as one. Live detection, scanning when the cluster changes, schedules and tags are set on the cluster's page, and every node follows them, nodes that join later included. A node can only be excluded. See The cluster is what you manage.
- One row per cluster on Hosts, with its nodes summed in it; the nodes are on the cluster's Nodes tab, with their live detection and agent versions.
- The cluster's Runtime tab: live detection on every node, with one stream of alerts naming the node each came from.
- Scan cluster and nodes: the cluster's scan checks the cluster and its nodes.
- Nodes that leave the cluster are removed 3 days later.
Before you upgrade
Back up: database schema 33. Hosts get whether a node is excluded and when it left its cluster, so 1.27 and earlier can't open the database afterwards. Back up first.
After you upgrade
The server moves each cluster's node settings to the cluster, so nothing that was on is turned off:
- Live detection: on for a cluster if it was on for any of its nodes, and then on for all its nodes. If you had turned it on for some nodes only, it's now on for all of them: use Exclude this node for the ones to leave out.
- Scanning when it changes: the same, from the nodes' Scan new containers automatically.
- Tags: the nodes' tags join the cluster's, and every node has them all. Notification channels and maintenance windows filtering on a node's tag now match the whole cluster.
- Schedules: a schedule that named a node now names its cluster, so it scans the cluster and all its nodes. Untick the cluster in the schedule, or exclude nodes, to scan less.
Then on the dashboard:
- Hosts shows one row per cluster; open it for its nodes (Nodes tab) and their alerts (Runtime tab).
- A node's page says Follows CLUSTER's settings. Its switches, tags and live detection are changed on the cluster now; scripts that set them on a node through the API get
409, naming the cluster: set them on the cluster. - Settings → Live detection → Turn on for new servers no longer applies to new nodes: they follow their cluster's switch.
Nodes are tied to their cluster by the cluster agent's scan. Nodes of a cluster without a cluster agent stay hosts of their own, as before.
Kubernetes (Helm and Kustomize)
Nothing to change in the chart or the Kustomize base beyond the image. The node agents and the cluster agent work with 1.28.0 as they are, and need no upgrade for the new settings: the server keeps them. A node agent older than the others is marked on the cluster's Nodes tab.
Install script, packages and Docker Compose
Nothing else to do.
Known issues
- A cluster's vulnerabilities, malware and file integrity aren't combined across its nodes yet: each node's are on its own page, and the Hosts list shows the most on any one node.
- Live detection alerts name the node, but not yet the container or pod.