This release fixes a class of bugs that might cause DRBD resources to end up with replica differences, even though DRBD shows them as UpToDate. It happened to resources that got a replica added later, after creation time. It only affects thinly-provisioned volumes, and only if they never had a write request with a node absent and fully utilized peer slots. (in other words, which are still on the day-0 current UUID).
In large, dynamic deployments, approximately 5-10% of the existing resources are affected. The DRBD team will follow up with more details in the announcement of the next DRBD release, which is scheduled for the next few days.
Additionally, fixes from the master branch have been cherry-picked, see changelog for details.
NOTE
With this release we also announce that we will remove etcd database support with Linstor 1.34.0, until then you will have time to migrate to (SQL or k8s-CRD) databases.
Heads up: Linstor 1.34.0 will switch to java-21, so be prepared to have java-21 jre ready for the next minor upgrade
[1.33.2] - 2026-04-08
Changed
- Changed the mechanism how LINSTOR sets DRBD volumes as UptoDate
Removed
- WinDRBD specific patches regarding missing sha1 and hmac crypto algorithms
Fixed
- Fixed resource definition delete setting DELETE flag when resources are still in use
- Fixed retry deletion of already deleting resource (no longer fails if it is the second to last diskful resource)
- Satellite now properly merges remote nodes + their data (props, netIfs)
- Fixed potential ConcurrentModificationException when serializing properties
- Fixed race condition in updateController() that leads to uncompleted ChangedController answers (client timeout)
- Fixed bug that could lead to NPE when multiple cache layers are processed in one DevMgr cycle
- Fixed possible bug that could have caused orphaned DRBD resources being left on the satellite
- Fixed autounplacer not prioritizing autoplacer-settings violations over free space properly
- Fixed EventStreamClosed exception if toggle-disk a single diskful resource
- Fixed some issues with ReconnectorTask when multiple peers representing the same nodes try to reconnect
- Fixed property "Linstor/Drbd/auto-block-size": Now works properly on controller level
- Fixed that additional volumes now also run mkfs if configured