This release fixes some small deployment issues with 2.0.0, along with one new feature which allows cleaner shutdowns of nodes.
The issue with 2.0.0 is that all DRBD devices are using the suspend-io strategy when dealing with connection loss. This means that during shutdown, when the Satellite Pod is stopped, all DRBD devices will suspend all IO. This makes it impossible for systemd to later cleanly unmount all file systems. To prevent this, we install a special runtime unit in systemd that will forcefully make all DRBD devices secondary, making it possible to unmount the device.
Added
drbd-shutdown-guardinit-container, installing a systemd unit that runs on shutdown.
It's purpose is to rundrbdsetup secondary --forceduring shutdown, so that systemd can unmount all volumes, even withsuspend-io.
Changed
- Updated LINSTOR CSI to 1.0.0, mounting the
/run/mountdirectory from the host to enable the_netdevmount option.
Fixed
- HA Controller deployment requires additional rules to run on OpenShift.