Skip to main content

Updates

During the Tech Preview, read the release notes before upgrading: any release can change behavior.

Updating TwinStor​

Updating TwinStor does not stop your VMs: storage keeps running while the service restarts. On each host, one at a time:

root@twinstor-1: update
# echo "<published sha256>  twinstor-test-<version>.tar.gz" | sha256sum -c
# tar xzf twinstor-test-<version>.tar.gz && cd twinstor-test-<version>
# ./install.sh
# twinstor upgrade-apply

twinstor upgrade-apply restarts the service only when it is safe (both copies in sync, the other host healthy), and returns once this host is healthy again. Then do the second host.

  • Never upgrade both hosts at the same time. upgrade-apply refuses on the second host until the first one is healthy.
  • twinstor upgrade-apply --check tells you whether an upgrade can run now, without changing anything.
  • No need to move VMs or turn HA off.
  • Finish with both hosts on the same version. If they stay on different versions, an alert is raised.
  • After some upgrades, twinstor status shows a config: line. This is safe to leave as it is. To clear it, in a maintenance window, run twinstor regenerate-config --persist-only --yes on one host, reboot it, wait for ready, then do the other.

From 0.7.1 to 0.7.2​

  • The first upgrade-apply sets the pool's default HA timeout if the pool has none (120 seconds fate-shared, 90 arbitrated), so turning HA on without a timeout, from Xen Orchestra or xe, no longer falls back to 60 seconds.
  • Log lines use plain names instead of internal tags. If you filter TwinStor logs on the scenario field, update the filter.

From 0.7.0 to 0.7.1​

Upgrade both hosts back to back: the fate-shared fixes only apply once both run 0.7.1.

From 0.6.x to 0.7.x​

Upgrade both hosts back to back, with no network work in between. Fate-shared freeze mode starts once both hosts run 0.7.0 or later.

Updating XCP-ng​

Use a Rolling Pool Update: turn HA off, then update and reboot one host at a time, then turn HA back on. Xen Orchestra's Rolling Pool Update does all of this for you, and nothing TwinStor-specific is needed.

Never update and reboot a host with HA still on. When turning HA back on by hand, use the command from Administration.

Updates with a new DRBD kernel module​

These need a reboot, so follow the Rolling Pool Update procedure above. The release notes say when this applies.