StorPool 19.3 release notes

Release 19.3, more details here.

19.3 revision 19.01.2592.cf99471bd

Released 25 Dec 2021

storpool_stat: additional data collected for more comprehensive monitoring alerts:
- modules tainted state from /proc
- free space on all filesystems
- lshw json output
- lspci with kernel modules output
storpool_ctl:
- just warns when the storpool_bd module could not be reloaded
- the status option now shows all alerts for kernel module versions
storpool_server:
- resets wbt_lat_usec parameter in /sys to 0 for all StorPool data disks
- fix for more than one disk with the same disk ID
storpool_mgmt/CLI - adds global max recovery request setup (see Local and remote recovery)

The new maxLocalRecoveryRequests and maxRemoteRecoveryRequests CLI parameters allow you to set the number of parallel requests to issue while performing local or remote recovery. These parameters replace the SP_NORMAL_RECOVERY_PARALLEL_REQUESTS_PER_DISK and SP_REMOTE_RECOVERY_PARALLEL_REQUESTS_PER_DISK configuration options, which are now deprecated.

19.3 revision 19.01.2571.5eb9133c9

Released 16 Dec 2021

storpol_bridge: use BIC for TCP connections to achieve higher throughput on high-latency links

19.3 revision 19.01.2562.7b73fb02d

Released 15 Dec 2021

storpool_server: fixes for a race-condition during re-balancing operation and system volumes aggregation that leads to abort.
Fixes for balancer tool for re-balancing a cluster with applied overrides.

19.3 revision 19.01.2545.66f61a9cd

Released 08 Dec 2021

Adds new features in storpool_volumecare version 1.21 (Change history).
Adds disk_init_helper - a set of tools will assist the setup with the initialization of NVMes, SSDs, PMEMs, and HDDs as StorPool data drives. For detailed usage and examples, see Storage devices
storpool_hugepages: now retries reservation

19.3 revision 19.01.2539.30ba167e1

Released 29 Nov 2021

Fixes for the improved handling of large volume and volume/snapshot chains (see 19.3 revision 19.01.2401.48d842f1d), related to re-balancing performance in clusters with HDD-only pools.
storpool_abrtsync: fix for service starting before new vmcore-dmesg file permissions in /var/crash get updated.
Initial support for persistent memory devices used for journal; for details, see Persistent memory support.

19.3 revision 19.01.2465.6f77d00cd

Released 11 Nov 2021

Critical fix for the monitoring agent that would miss a total API outage (no active API running in a cluster).
storpool_mgmt - fixes case with a destroyed snapshot on an ejected and forgotten disk.
Adds support for all 5.11+ kernels (up to 5.13 for Ubuntu 20.04 LTS).

19.3 revision 19.01.2456.977787e12

Released 02 Nov 2021

Fixes in storpool_mgmt related to the migration to the improved internal metadata structure (see 19.3 revision 19.01.2401.48d842f1d).
Initial build and internal tools support for AlmaLinux and RockyLinux 8.4.

19.3 revision 19.01.2439.425b19e8d

Released 26 Oct 2021

storpool_mgmt:
- fixes a known VolumeGetStatus issue that may cause the API to abort/restart.
- fixes entering maintenance mode with remote volume export
Python API bindings: adds the volumeStatusQuick API call.
storpool_stat: collects inventory less often, reduces CPU usage.
storpool_abrtsync: proxy support.

19.3 revision 19.01.2426.40f100363

Released 08 Oct 2021

Python API bindings 7.0.0: code cleanup, adds the globalId member to the AttachmentDesc structure.
Python API bindings: adds the iSCSI targets in the ClusterStatus response.
Python spopenstack helper 3.1.0: code cleanup, changes the locking strategy.
Drops some unneeded work that the storpool_ctl tool did at each invocation.
Fixes a minor problem with collecting internal diagnostic information in an edge case with a specific version of the confget tool installed.

19.3 revision 19.01.2401.48d842f1d

Released 29 Sep 2021

Improved handling of large volume and volume/snapshot chains
The internal metadata structure is updated for more efficient metadata operations. The upgrade will provide:
Significantly improved flexibility for clusters with a large number of volumes/snapshots and chains of volumes or snapshots with sizes larger than 1TiB.

Better overall relocator performance.

More efficient object use for large snapshots.

Improved overall snapshot delete cycle.
StorPool support team handles the upgrade to meta V2. The upgrade is in the internal data structures, and will be enabled once all services in the cluster are running the latest version.

Once upgraded, the size of the disks sets table can be updated for each volume (or snapshot) to up to 32K. For example, a very large volume filled with data can be updated to a larger disk sets table size, providing more granularity when re-balancing its data. This set of features will be very helpful when a cluster has a small set of disks (the usual for a backup only cluster) or disks with different sizes, typical for clusters in process of expansion or hardware being upgraded/replaced.

Note

Note that expansions of the volume/snapshot disk table with a larger size will require a re-balancing operation, though shrinking will be applied immediately by the relocator.

There are few small changes in the output for some commands after the upgrade to meta V2 is completed, for example, the storpool task list command output will now flag all operations related to a change in the volume or snapshot policy with balancer (example output in Tasks).
Adds fix for storpool_stat service to properly cleanup processes after data collection (no leftover zombie processes).
Adds support for up to 16 server instances per node for high-performance use cases.
Fix for extending active maintenance.

19.3 revision 19.01.2333.b24f05090

Released 21 Sep 2021

Adds support for Connect-X6 DX/LX, Connect-X7 network interface cards.
Adds support for Intel P5510 NVMe devices.
Adds fix for VolumesGetStatus API call showing volumes/snapshots as degraded.

19.3 revision 19.01.2321.a82720c45

Released 10 Sep 2021

Adds fix for balancer to correctly handle overrides.

19.3 revision 19.01.2319.27c976e3e

Released 07 Sep 2021

Simplify/clarify all messages from kernel modules.
Log the name of any process that opens a StorPool block device for writing.
Add an option to control the logging of opening of StorPool block devices (Logging for opening or closing of StorPool devices).

19.3 revision 19.01.2318.10e55fce0

Released 30 Aug 2021

Speeds VolumesGetStatus up in clusters with many volumes and snapshots.
Adds initial build and internal tools support for RHEL 8.4.
API access settings can now be overriden for all tools using the Python bindings through environment variables.
Compatibility improvements for NICs using the bnxt_en driver.
Volumes or snapshots with constraints violated are now visible in the output of storpool volume status; for details, see Volume status and Reusing server on disk failure.
The AttachmentsList API call and the storpool attach list CLI command now show the globalId of the attachment.
The storpool_abrtsync service now uses HTTP/HTTPS for all transfers by default, instead of rsync/SSH.
Adds fix for handling TRIM commands for Windows-based initiators.

19.3 revision 19.01.2268.656ce3b10

Released: 06 Aug 2021

Adds support for Samsung PM9A3 (MZQL27T6HBLA-00A07).
Adds a change in the way volume overrides are loaded.
Includes multiple reliability and stability fixes around storpool_iscsi, storpool_bridge, and storpool_mgmt services.

19.3 revision 19.01.2216.14df508f2

Released: 27 Jul 2021

Improves the API documentation structure.
Improve the storpool_block service’s handling of hot-plugged memory blocks.
The snapshot delayed delete option is now on by default, as it is being used by the ONE integration for safety when deleting virtual machines.

19.3 revision 19.01.2199.8700d0744

Raised buffer limits to handle high-bandwidth/high-latency inter-cluster connections.
Adds initial build infrastructure support for ROCKY8/ALMA8 Linux.
Includes ioatdma to blacklisted modules so that Intel IOAT acceleration could be enabled on supported systems.
storpool_cg changes:
- Adds auto-detection for the optimal number of CPU cores on server-only systems.
- Now supports changing the default machine.slice with a custom name (used for Kubernetes automation).
Improved storpool_bd kernel module logging.

Previous release

19.2 revision 19.01.2173.0ca29830b