Upgrade
1
See what's new
Each version’s changes are in the changelog and on GitHub releases.
2
Upgrade the chart
--version, Helm installs the latest. The image’s tag follows the chart unless you set image.tag.3
Check it started
Lumovi … shows … names the version that’s running. The log’s lines of JSON, with "type":"lumovi.audit", are its audit events.What people see
A new pod signs nobody out, with one exception where single sign-on passes people’s tokens on: Lumovi keeps sessions, and the AI assistants people allowed, in a Secret the chart makes (auth.keepSessions), which upgrades leave as it is. Open pages reconnect to the new pod by themselves, and carry on. AI assistants open a new connection; those that pass their person’s own token on wait until that person opens Lumovi again. See Across restarts. With auth.keepSessions: false, a new pod signs everyone out, and open pages show the sign-in page.
The same goes for any change that replaces the pod: changing a setting, or the plugins everyone sees.
For the same reason the chart runs one replica, and refuses more while AI assistants or the audit history’s volume are on, as they are by default. See
replicaCount in Helm values.When the client secret changes
A change to a Secret’s contents doesn’t replace the pod. Lumovi reads the single sign-on client’s secret from an environment variable, when the pod starts, so it keeps the old one until it restarts. That’s the case when you changeauth.oidc.clientSecret and upgrade (the chart changes its Secret, not the pod), and when you change the Secret that auth.oidc.existingSecret names.
Restart Lumovi so it reads the new secret:
fleet.agentsSecret names, is read the same way, when the pod starts: restart Lumovi after changing it too. Its kubeconfig, in fleet.kubeconfigSecret, and the Secrets it reads with fleet.secrets are read again by themselves, without a restart.
A fleet
Upgrade a fleet’s Lumovi as above, or, where it runs from the image alone (with Docker, or on a platform like Sevalla), by changing the image’s tag.- Its agents connect again by themselves. Stopping, Lumovi closes their connections, and each agent tries again a second later, then less and less often, until the new one answers. Lumovi’s log names each agent’s version as it connects, like
The agent of private-eu connected (Lumovi 1.23.2). - Members and agents are releases of their own, in their own clusters. Upgrade each there, the same way, with its own values or
--reuse-values.
Roll back
Helm keeps the previous revision, so going back is one command:helm history lumovi --namespace lumovi lists the revisions, and helm rollback lumovi <revision> --namespace lumovi goes back to a particular one.
Changelog
What changed in each version.
Helm values
Every setting the chart has.