v0.6.6

What's Changed

New Features

Install SW — mc-application-manager install screen in the console

• Install SW opens the mc-application-manager install screen in an iframe popup with the target already set, instead of navigating away and losing the selected infra/node
• Available for a single node and for a whole node group: on the Server tab for the selected node, and on the Group tab's Node List card for every node in the node group
• Enabled only for running nodes: a node must be Running, and a node group needs at least one running node. When disabled, the reason is shown on hover and as a toast on click
• Console terms are translated to the application-manager contract (infra/node → targetType=VM, mciId, vmId); the install screen rejects any other parameter
• Origin-checked postMessage handshake: after IFRAME_READY the console sends the token, workspace and project to the application-manager origin, and only accepts result messages whose source, origin, type and request id match

SSH Keys — private key available after creation

• The private key is shown in the detail panel at any time, masked as ****** by default, with Show/Hide and Copy. It used to be visible only right after creation, so a key reopened from the list showed nothing
• Copy works without revealing the key on screen
• Clearer empty states: Not available when there is no private key, and Not provided by the CSP instead of None when the CSP returns no public key (AWS key pairs)

Bug Fixes

• Deleting a checkbox-selected K8s cluster or Infra did nothing: a checkbox click does not fire rowClick, so the current selection was never set — the K8s delete request was not sent and the Infra delete request went out with an empty id. A single checkbox selection now updates the selection the same way a row click does, and deleting with several K8s clusters checked asks the user to pick one
• Node Delete on the Group tab sent an empty nodeId or deleted the node selected on the Server tab — it now deletes the node chosen on the Group tab, and empty ids are blocked before the request
• Infra/Node deletion left CSP instances running: deletion used option=force, which removes only the CB-Tumblebug record. It now uses option=terminate, so the CSP resources are terminated as well
• AWS cluster creation in Expert mode always failed: only one subnet could be selected and AWS EKS requires subnets in at least two availability zones. The subnet field becomes a multi-select for AWS, shows each subnet's zone and CIDR, checks the zone count before sending, and sends every selected subnet in subnetIds
• Running nodes showed an empty public IP: the node detail view discarded the GetInfraNode response because of a wrong comparison, and CB-Tumblebug 0.13.4 stores an empty publicIP right after Resume. The response is now applied, and an empty IP on a running node is filled from the CSP with GetInfraNode option=accessinfo
• Kubeconfig requested for clusters still being created: GetK8sClusterKubeconfig was called before the cluster status was known. It is now requested only for Active clusters
• "processing…" progress toasts never closed: the toast was registered under a fixed id with the wrong arguments, so closing it by request id did nothing. Progress toasts now use their own id and label and close on success and failure, and the loader is closed before a 401 token-refresh retry
• Sidebar active state: iframe menus did not expand their parent dropdown or mark the item active, and local menus highlighted every sibling item under the same parent
• Remote command failures in the Node Terminal showed only (No output): PostCmdInfra returns the failure reason in error (a string), but the terminal only checked err. It now reads error first and still accepts the older err object
• Projects added after login disappeared from the top Project selector after navigating: the project list is read from a session cache filled at login, the refresh button did not query again while a cache existed, and the list fetched on workspace selection was not written back to the cache. The refresh button now reloads the workspace/project list and updates the cache (the existing cache is kept if the request fails), and the list fetched on workspace selection is cached as well. Page loads still read only the cache, so no extra API calls are made

Dependencies

• mc-iam-manager v0.6.4: conf/api.yaml now records mc-iam-manager version 0.6.4 (was 0.6.3)

Full Changelog: v0.6.5...v0.6.6