Uninstall

This page shows how to uninstall node, ctrl, and the two gateways, and what data stays on the hosts and in external storage afterward. To reinstall or upgrade a package, see Reinstall and upgrade.

NotesCopy Anchor Link

• The package scripts refuse to replace or remove the package when the service is still running, the systemd state is unknown, the configuration is missing but leftover state exists, or the maintenance check fails. After you fix the problem, retry the same operation.
• Both remove and purge delete only the package files and the service's enablement links. The data listed in Data left behind is kept. To destroy it, the deployer does so separately.

Order across componentsCopy Anchor Link

To remove a whole deployment, uninstall the components in this order:

1. The E2B gateway and the Exposure gateway, so that no new external requests arrive. See Uninstalling the gateways.
2. Each node. See Uninstalling a node. Draining, cleaning up, and retiring a node all go through ctrl.
3. ctrl, last. See Uninstalling ctrl.

Uninstalling a nodeCopy Anchor Link

1. Drain the node so that ctrl places no new Workspaces on it. Drain does not clean up any data.
2. Delete or export all Workspaces, detached Volumes, and retained layers on the node. kibo workspace list --node node-a --all -q lists its Workspaces.
3. Wait until the node reaches DRAINED, then retire it with kibo node retire. ctrl then no longer counts it as a member of any storage pool, and the node ID cannot be enrolled again. Retire while the node service is still running: a stopped node is no longer DRAINED, and retire then accepts it only after kibo node isolate.
4. Stop the service, run the maintenance check, and purge the package.

Drain the node and list what is left on it:

kibo node drain node-a
kibo workspace list --node node-a --all -q

After the cleanup, wait for DRAINED, retire the node, and remove the package:

kibo node drain node-a --wait --timeout 10m
kibo node retire node-a
sudo systemctl stop kibotos-node.service
sudo kibotos-node check-maintenance --config /etc/kibotos/node/config.toml --operation remove
sudo apt-get purge kibotos-node

The package scripts run the same --operation remove check, disable the service, and end the data-plane and egress processes. For retire and what happens to a pool that only this node reached, see Nodes.

Uninstalling ctrlCopy Anchor Link

Uninstall ctrl after all nodes and gateways. Stop the service, then purge the package; the package scripts disable the service:

sudo systemctl stop kibotos-ctrl.service
sudo apt-get purge kibotos-ctrl

Reinstalling node after uninstallCopy Anchor Link

After uninstall, /etc/kibotos/node and /var/lib/kibotos/node still exist, but the kibotos-node binary has been removed, so the install script has no binary to read these records with. When you reinstall:

• If any kibotos-vmm-* units remain, the installation is refused and lists these units. After you confirm they are no longer needed, stop them (run systemctl reset-failed for any that are still listed), then retry the installation.
• If there are no VMM units, the installation continues, keeps the existing configuration, identity, and records, and tells you to check before starting.

After the installation finishes and before you start the service, check the retained state with the newly installed binary:

sudo kibotos-node check-maintenance --config /etc/kibotos/node/config.toml --operation upgrade
sudo systemctl enable --now kibotos-node.service

Keeping the identity works only for a node you did not retire: a retired node can never register again. If you do not plan to keep the existing identity, delete these two directories before installing, and enroll again by following Install node. After a retirement, use a new node ID.

Uninstalling the gatewaysCopy Anchor Link

Stop the service first, then remove the package. The package scripts disable the service:

sudo systemctl stop kibotos-e2b.service kibotos-gw.service
sudo apt-get remove kibotos-e2b kibotos-gw

Data left behindCopy Anchor Link

Neither remove nor purge deletes the following. Delete what you no longer need yourself:

• A shared pool on NFS may still hold other nodes' objects. Delete its directory only after every node that declares it is gone.
• Keep pki/ and credential-keys/ if you may reinstall ctrl on the same database. Without pki/, all nodes and Operators must enroll again. Without credential-keys/, the caller credentials stored in the database cannot be decrypted.