Release notes

StorPool 21.0

21.0 revision 21.0.1041.42d32fec3

Released 22 Nov 2024

Disable needrestart for all StorPool services in Ubuntu 24.04 LTS so that it does no longer restart services upon a new installation.
Added a fix for disk aggregation control bug in storpool_server that could lead to increased latency and degraded performance.

21.0 revision 21.0.971.f7fba2495

Released 23 Oct 2024

Introduced a new set of lower default network event timeouts that reduces the max latency outliers in some networks supposedly during switch buffers congestion or overload.
Added a fix that is speeding up rebalancing in EC clusters with irreversibly failed disks or nodes.
Added a more memory efficient logging facility, which should now no longer trigger alerts for the storpool_mgmt memory cgroup with enough cached logs.
Added a fix in storpool_iscsi to correctly handle net-masks that are not /24, /16 or /8 networks.
Added fixes in multiple places in the upgrade tooling to fail early to prevent updating the same node more than once.
Added a fix to properly rotate the logs from the storpool_volumecare service.

21.0 revision 21.0.956.a36a6ff17

Released 12 Aug 2024

Updated tooling (the abandon-down-drives tool) part of the procedure for bringing a cluster back in running state faster, after the following conditions:
- An irreversibly failed drive.
- A power-loss event.
Added a fix for the storpool_havm service to keep a VM running until a stable API is back up and running.
Updated VolumeCare to version 1.29.3; for details, see Change history.
Added improvements in the storpool_server service:
- Added a fix for stuck read from an erasure coded volume with disks missing on more than one node with erasure coding (EC) 8+2.
- Added a fix for a complicated case with aggregating entries with interleaved trims and data.
- Introduced the ability to delete and dematerialize newly created snapshots with a down disk.
- Introduced fixes for slower/stuck re-balance operations in EC clusters during redundancy restore.
- Added a fix for a crash during initial EC conversion.
Added a fix in storpool_initdisk to recognize disks with overwritten final 4k.

21.0 revision 21.0.841.983f5880c

Released 12 Aug 2024

storpool_mgmt: added a fix for crash when creating a volume from snapshot, and the volume is larger than the snapshot.
storpool_mgmt: added configurable space aggregation algorithm parameters in order to lessen the latency impact from space aggregation when a large amount of data gets deleted from a disk (details).
storpool_bd: added the SP_BD_NOSLEEP option to disable push-back on full IO queue.

21.0 revision 21.0.809.72cdd84cd

Released 08 Jul 2024

Added support for Ubuntu 24.04 LTS.
Added support for 6.8 line of kernels under Proxmox Virtual Environment 8.
storpool_mgmt: better handling of relocation/balancing in EC clusters.

21.0 revision 21.0.691.da4ac3daf

Released 26 Jun 2024

Performance improvements with mlx5_core based NICs with lower interfaces speeds (2x10GE) on Linux kernels 5.15 and later.
VolumeCare update to version 1.29.2; for details, see Change history.
storpool_mgmt: handle case with hanging reads in multicluster for snapshots that are recovering from remote (backup) cluster.
storpool_block, storpool_iscsi: fix crash on queue full and requests for an idle volume.

21.0 revision 21.0.670.8737a7719

Released 07 Jun 2024

Disable VF creation on Mellanox ConnectX-6 Lx/Dx NICs with recent firmware due to multicast storm issues.
Blacklist acpi_pad module, which sometimes is resulting to very difficult to debug IO stalls of the storage system.

21.0 revision 21.0.647.9a4154290

Released 17 May 2024

Fix in storpool_mgmt showing lower than real stored size for volumes and snapshots.
Fix in storpool_iscsi to work against TCP Window Scaling = 1 (default in Ubuntu 24.04).
Fix in relocator to not overflow disks in erasure-coded clusters during rebalance.

21.0 revision 21.0.606.609fab857

Released 02 May 2024

Fix in storpool_mgmt to not falsely deplete objects on down disks.

21.0 revision 21.0.576.6d56902e2

Released 23 Apr 2024

Erasure Coding:
- Improved relocator operation with erasure coding when re-balancing out ejected disks.
- Improvements in maintenance set logic while there are erasure coding tasks in progress.
Added support for the WD Ultrastar DC SN650 series drives.
Updated capabilities of storpool_cg cgroups configuration tool to keep the last successful configuration in a continuously automatically updated configuration file. The update also removes the support for the -D / --dump-configfile option for writing the configuration to file. For details, see Configuration options and parameters.
Added -E/--expect option for storpool_cg print. It can be used for printing the expected memory usage. For details, see Verifying machine’s cgroups state and configurations.

21.0 revision 21.0.359.dfeb4244c

Released 05 Apr 2024

Fix for storpool_iscsi causing sessions to reconnect upon creating a snapshot.
VolumeCare update to version 1.29 (more info at Change history).

21.0 revision 21.0.330.28fa27b64

Released 07 Mar 2024

Reduced memory requirements for the storpool_iscsi service.
The storpool_beacon service no longer supports compatibility with releases earlier than 20.0
Removed hardware acceleration support for bnx2x based network interface controllers.
Improved latency of operations in multi-cluster environments while there are active transfers.

21.0 revision 21.0.318.051ec1d80

Released 20 Feb 2024

storpool_mgmt: handle some cases of stale snapshots during multicluster migrations
Performance improvements with storpool_iscsi and faster network (25+ GE).
Added support for 6.5 line of kernels.

21.0 revision 21.0.288.e52a50c1a

Released 28 Jan 2024

Fix for mlx5 driver for hang due to queue overflow.
The update tooling gets all services back up even in case the command is interrupted.

21.0 revision 21.0.266.455001e21

Released 16 Jan 2024

Move some network tuning sysctls to be activated only on bridge nodes.
The VolumeFreeze API call is in process of deprecation, the storpool CLI will start showing deprecation warnings, until this is completed.
Fix for mlx5_core based NICs for hanging traffic due to incorrect signalization handling.
Adds support for SK Hynix PE8110 E1.S series NVMe devices.
Improved the storpool_tree tool. For details, see StorPool tree.

21.0 revision 21.0.242.e4067e0e4

Released 4 Jan 2024

Hardware acceleration related fixes for the mlx5 driver with an older CentOS 7 kernel.
VolumeCare update to version 1.27.1 (more info at Change history).
New balancer options for better handling of degraded states: --ignore-down-disks and --empty-down-disks. For details, see Options
New helper tool storpool_capacity_planner for better planning of hardware upgrades for different erasure coding schemes. For details, see StorPool capacity planner.

21.0 revision 21.0.216.701e0f8e0

Released 15 Dec 2023

Erasure Coding:
- Performance and stability fixes with very large volumes.
- storpool_mgmt: fix for handling re-balancing a cluster with active erasure coding with a forgotten disk.
VolumeCare update to version 1.27 (more info at Change history).

21.0 revision 21.0.94.f7de41582

Released 25 Oct 2023

Improved API responsiveness during high load while performing remote recovery into a local hybrid pool.

21.0 revision 21.0.75.1e0880427

Released 13 Oct 2023

Introduced new feature: Erasure Coding

StorPool introduces a new redundancy mechanism called erasure coding, which can be applied on systems based on NVMe drives. It can be used as a replacement for the standard replication mechanism offered by StorPool, which would reduce the overhead and the amount of hardware needed for storing data reliably.

Migration to an erasure-coded scheme happens online without interruption in the service, and is handled by the StorPool support team for the initial conversion of a running cluster.

For more information on using erasure coding and what advantages it provides , see Erasure Coding.
Added support for Debian 12 and Proxmox Virtual Environment 8.
All services local to a node are now able to detect if their storpool_beacon service is hung.
Introduced an additional check in the update_rdma tool that the kernel modules to be reloaded exist for the running kernel.
Prevented StorPool services from using UDP port 27489 in order to prevent collisions with other services running on the node or in the storage network.
Added support for the Micron 9400 PRO.
Reduced the amount of logging from all StorPool services.