Changelog: what’s new in Clusterward

A log viewer merges all instances of a service by time, with search, “Errors only”, download and the previous run of a crashed container.

CPU, memory and instances over 24 hours, 7 or 30 days; uptime checks test your domains from the outside and report outages.

CPU and memory on a 0–100 % axis of the limit, for Helm chart services too. Ranges from 15 minutes to 30 days, and a large view with the current, average and peak value.

Notifications, invitations and password e-mails show the sending operator’s domain in the sender name, the subject and the footer, along with its company details and imprint and privacy links.

Under Operations → SQL console, run one statement against a PostgreSQL or MySQL database: with its own read-only login, in a read-only transaction, up to 60 seconds and 1,000 rows by default. Every query is in the audit log, and the console has a permission of its own.

Automatic snapshots, backups of individual databases with download, and restore into a new database alongside the running one.

Bring back a volume from a snapshot, with a safety snapshot taken first; the disks that come with a Helm chart get snapshots too.

Deploy and restart from GitHub Actions or any script, with its own role, scope and expiration date.

New users set their own password via an invitation link, recovery codes replace a lost phone, and attacks on sign-in are blocked.

Every change searchable in the cockpit; the entire configuration as a JSON file, without passwords and keys.

Every environment shows whether it is isolated: then only the ingress controller and its own pods get in, and the nodes’ metadata service is out of reach. Isolating takes one click, for new environments and pipeline environments right after the first rollout; System → Security lists all environments. It is offered only where every prerequisite is met.

An isolated environment can be opened again. When an ingress controller is added or removed, the policies of every isolated environment on the cluster adjust themselves.

Entering keys and tokens, destructive actions and writing Helm charts now need the manage level. Everyone grants roles, scopes and API tokens only within their own rights, and the application scope now also covers tenant runs, backups and bucket keys.

Secrets can no longer be read after saving and, if you choose, are stored in your project’s Secret Manager, in folders per app, environment and service.

A new version in Secret Manager is detected and, if you choose, rolled out with a restart; selectable per cluster via the External Secrets Operator.

A secret store can get a read-only key that External Secrets uses to deliver values into your namespaces. The store page shows where one is still missing.

“Switch class…” shows for every host before it starts who moves the DNS record and what happens to the certificate. Clusterward moves its own and adopted records itself, detects the move behind the Cloudflare proxy too and lets you abort a running switch.

Registered zones are compared with Cloudflare every hour. Existing records can be adopted: Clusterward then changes them like its own but never deletes them. The Domains card reads every host live.

Envoy Gateway as a second ingress controller; switching the ingress class runs both load balancers in parallel until DNS has moved over.

Timeouts, retries, rate limits, upload size, CORS, sticky sessions and basic auth per service.

When a host behind Cloudflare’s proxy redirects, Clusterward writes a redirect rule into the zone on every deploy: one 301 to https, deeper pages keep their path. Targets may now carry a path, such as example.com/de. The zone token needs the “Dynamic URL Redirects: Edit” permission for this.

If a chart has an ingress.mode switch, moving between nginx and Envoy runs as a cutover for Helm services too.

A wildcard certificate issued by Clusterward for *.example.com also covers example.com.

Allow your own address for the Kubernetes API with one click – without the Scaleway console.

The cluster identity reads and writes secrets only in your environments’ namespaces, which run with Pod Security “baseline”; a policy on the cluster keeps it that way. Builds push with a key that may only use the registry, and admin access to a cluster lasts one hour. Switch existing clusters over on the cluster page (Kubernetes 1.30 or newer).

KEY=secret:value in an environment block’s env template becomes a write-only secret of the tenant’s service, in Scaleway Secret Manager if you like. Workload blocks create chart secrets before the first rollout, and env templates create secrets per service without holding any values themselves.

A pipeline’s domain block can name an ingress class, so the record points at the right load balancer.