You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Patch Changes
197c333: Add nipa completion --install, which saves the tab completion script where bash, zsh or fish loads it, so the shell doesn't run nipa each time it starts. nipa rewrites the saved script when its version changes, so new commands complete after an update. nipa completion without a shell uses the one in $SHELL.
47797cd: nipa server ls marks the servers that a Kubernetes cluster made, such as k8s-control-plane-1 (Kubernetes master), from the Magnum tags the portal reads too. --json gives each server kubernetes, which is null for a server you made, or the cluster's clusterId and the node's role. The troubleshooting page explains why nipa os coe and other commands can't find their service on production.
587c9ac: Give production's Space API only to profiles on production's Keystone. In 0.1.4, a staging profile got production's Space API, which refused its token, so nipa server ls said the session had expired. nipa now drops that URL from such profiles. nipa profile add --space-url takes the Space portal URL, such as https://portal-stg-epc.nipa.cloud, saves its /api, and checks that it answers. With another Keystone and a terminal, profile add asks for the portal URL. Without one, server ls, db ls, lb ls and ip ls say the profile has no Space API URL before they ask for a login. A profile that 0.1.4 saved with a portal URL gets its /api when nipa loads it, and a Space API URL that answers with a web page gets an error that says so.