Community bug-fix release.
Upgrading: the node DaemonSet and the csi-provisioner arguments changed, so standalone
deployments must re-apply deploy/truenas-csi-driver.yaml. Bumping the image tag alone
does not deliver the iSCSI connector, iscsiadm, or topology fixes below. Operator
deployments pick the changes up on the next reconcile.
Fixes:
- #55: iSCSI and NVMe-oF connector state is kept on the host, so a csi-node restart no longer strands staged volumes. Unstage now reports failures instead of silently succeeding, and stage re-mounts a staging path left read-only or unresponsive by a storage-side outage rather than handing it to pods.
- #56: the host's iscsiadm is resolved on PATH instead of a hardcoded /usr/sbin/iscsiadm, so iSCSI works on hosts that install it elsewhere (Talos uses /usr/local/sbin). iSCSI failures now include the command's stderr instead of a bare exit status.
- #57: the driver no longer reports accessible topology or advertises VOLUME_ACCESSIBILITY_CONSTRAINTS. TrueNAS storage is reachable from every node, so volumes are never pinned by node affinity.
- #58: a repeated CreateVolume for a snapshot- or volume-backed request returns the requested content source, so the provisioner no longer deletes a freshly restored clone.
- #54: the client detects an expired appliance session and reconnects, ending the ENOTAUTHENTICATED failures that began after long controller uptime.
- #53: corrected the k3s kubelet path in the README. k3s uses the standard /var/lib/kubelet.
Images:
- quay.io/truenas_solutions/truenas-csi:v1.2.0
- quay.io/truenas_solutions/truenas-csi-operator:v1.2.0
- quay.io/truenas_solutions/truenas-csi-operator-bundle:v1.2.0
Full changelog: v1.1.2...v1.2.0