[1.34.2] - 2026-07-24
Fixed
- Fixed the
zfs rollbacksnapshot rollback strategy reporting success as soon as the satellites confirmed the resource re-activation; it now waits for the rolled-back DRBD resources to actually become ready again, like resource creation and snapshot restore do - Fixed a changed DRBD auto verify algorithm not being deployed to already connected satellites when recomputed after a node full-sync, leaving nodes with different
verify-algsettings and forcing their DRBD connections into StandAlone - Fixed the controller not being able to execute
journalctlto collect systemd logs by adding it to thesystemd-journalgroup - Fixed drbd-proxy disable not releasing the target-side proxy TCP port, leaving it permanently allocated in the resource connection and the target node's port pool
- Fixed resource-connection DRBD Proxy ports not being re-reserved in the node TCP port pools on controller startup, so a restart could hand out a port still in use by DRBD Proxy
- Fixed the failure path of DRBD port assignment leaving previously reserved ports unreserved in the pool
- Fixed the resource definition staying marked as down on the controller after a successful snapshot rollback using the ZFS rollback strategy (the cleared flag was never committed)
- Fixed incorrect usage of SO_REUSEADDR introduced in fb5a9ac. Plus, IP address is no longer ignored (i.e. not using 0.0.0.0) when testing ports.