Full-Cluster Upgrade

If work, etc, and log are inside the install tree, move them outside before deleting the install tree. (If they are already outside the install tree, no action is needed.)

Replace the contents of the install directory with the unpacked new ZIP.

Copy gridgain.properties from the new install’s etc/ directory over the preserved file in your etc/ directory, so that the version metadata matches the new binaries. This step is specific to ZIP installs; the DEB/RPM and Docker procedures handle the equivalent automatically.

With the node already stopped from phase 2, run the standard package-manager upgrade command for your distribution. The package upgrade preserves any existing vars.env and gridgain-config.conf files in a backup/ subdirectory under the configuration directory, ensures the install, configuration, log, PID, and work directories exist with the correct ownership, and registers the service unit (systemd or Upstart) for the new binaries. The persistent work directory is left untouched, so existing cluster state is preserved.

Do not start the service yet — phase 4 starts every node together.

Update the image tag in your run command, Compose file, or Kubernetes manifest to the new version. The container itself is started in phase 4, against the same persistent volume that holds work, etc, and log.