Releases: LINBIT/linstor-server
Releases · LINBIT/linstor-server
v1.29.1
[1.29.1] - 2024-09-24
Changed
- LVM cache is now time-based, configurable by new property StorDriver/SizesCacheTime
- sedutils-cli calls will now resolve given paths before using them (as sedutils-cli only allows /dev/nvme*)
- resource-definition properties are now included in drbd connection net section
- Attempt to delete LV during DELETE flag even if SkipDisk is enabled (suppressing error reports)
- SOS: Additionally collect journal-based dmesg from multiple previous boots if necessary
Fixed
- storage-pool list reporting offline nodes even tough everything is connected
- SED setup using same password for all drives
- Fixed stuck deletion of resource during failed / failing resize
- Fixed failing rsc-dfn-update-chain when updating rsc-grp
- Fixed rare race-condition during snapshot create
- Fixed dateformatting with offsets
v1.29.0
[1.29.0] - 2024-07-31
Changed
- Improved responsiveness: Allow some API calls to be executed concurrently on the satellite
- Added a cache for lvs and vgs/pvs commands, reducing the need to call them
- Changed logback log format to include logid and full timestamp
Fixed
- Snap, create: Fixed possible deadlock
- Incorrect resource definition already exists error message
- sos-report: Controller not always dumping full log content
v1.28.0
[1.28.0] - 2024-07-11
Added
- Autoplacer: Add --x-replicas-on-different option
- Resource delete: Add --keep-tiebreaker option
Changed
- Snapshots now cannot be created while SkipDisk is active
- Autoplacer: Added very small default weight for the MinResourceCount-strategy for better tiebreaker-distribution
- Added some more info logging and improved an error message
- BalanceResourceTask: Do not count skipDisk resources as diskful
Fixed
- SkipDisk: layers below DRBD now get ignored to prevent errors on the satellites
- ReconnectorTask,EventProcessor: Attempt to fix possible deadlocks while having many connections/requests
- Clone: Fix clone waiting if there is more than one diskless resource
- ToggleDisk: Fix toggle disk for resource spawned with diskles-on-remaining
- Backups: Ensure snapshots can't be stuck in either sending or receiving, which would render them undeletable
- PhassPhraseInit: Fixed uncommitted transaction
- Backup,restore: Check correct SnapVlmDfn property before recalculating AllocationGranularity
- FreeSpace: Fixed incorrect calculation of reserved space for thin volumes
v1.27.1
[1.27.1] - 2024-04-25
Added
- Support for java-21
Changed
- Freespace calc: Snapshots only reserve allocated size now instead of volume size
- resource-definition modify now, only warns if a satellite is offline
- AutoVerifyAlgo: allow to use algorithm name additionally to the specific implementation
- ZFS: try to determine default volblocksize
Fixed
- Deleting a remote can now be safely done even if there are active shippings to said remote
- MakeAvail: trigger update satellite if only FLAGS changed
- Resource delete: Don't check user set verify-algo if last resource deleted.
- If mkfs failed, resources have not been demoted(secondary) and couldn't easily be deleted afterward
- Fixed incorrect least common multiple call when calculating AllocationGranularity
v1.27.0
[1.27.0] - 2024-04-02
Added
- Allow users to provide their own encryption passphrase for volumes
- Allow users to provide their own encryption passphrase after cloning
- Allow users to change the passphrase for a luks volume definition
- Backup ship, restore and scheduled backups now have a --force-restore option
- Added more info logging
- Added "DrbdOptions/ExactSize" for migration purposes
Changed
- Default configuration directory for linstor-database utility (export-db and import-db) is now "/etc/linstor"
- Make-available will no longer return an error if an involved node is offline (warn instead)
Fixed
- rscDfn-props are no longer reset through --download-only backup restore or if it has rscs deployed
- Fixed snapshot-rollback more than once on a resource
- Fixed re-/ordering issues within the AutoSnapshotTask
- Fixed backup restore not working if one of the backup-parts already existed on the cluster
- Fixed K8s migrations to use correct generated DB table instances
- Deleted invalid SpaceHistory entries from K8s.
- Fixed left-over tiebreaker after evacuation of single DRBD resource
- Fixed negative result of an allocation granularity calculation
- Fixed infinite loop and limited range of power-of-2 calculations
- Workaround for connections where the attached peer object is missing
- Only show warnings of missing cgroup (v1) if the user actually tries to use it
- Fixed undelete resources (also remove volume DELETE and DRBD_DELETE flags)
v1.26.2
v1.26.1
[1.26.1] - 2024-02-22
Changed
- LVM: Removed trailing 'a|.*|' from LVM filter
- Do not write/check drbd resource files if nothing changed
- RG spawn: No longer display "TransactionList" when spawn fails
- Sp,Props: Do not trigger a DevMgrRun for certain property changes
- Tiebreaker: Allow in all even-numbered (>0) diskful and 0 diskless setups
- Improved error message when FullSync fails due to missing external tools
- SOS-report: add more information output (lvs, zfs list, client output....)
- ProblemReports now also get added to the error-database (i.e. will show in up "err list")
- Snapshots: allow taking snapshot if at least one diskfull node is online
Fixed
- Do not update satellites that are evicted
- Node/Restore: Fixed resource kept in inactive state until controller restart
- Node/Restore: Fixed node not connected on node restore
- Luks: Allow deletion without entered master-passphrase
- SysFs: Warn if /sys/fs/cgroup/blkio does not exist
- When evacuating or migrating an InUse resources, wait until the resource is no longer in-use to remove resource.
- Restoring a pre-SP-mixing-backup into an SP mixing scenario
- SpMixing: ZFS <-> ZFS_THIN is no longer considered as mixed SP
- Encryption: don't create error-reports for user errors
v1.26.0
[1.26.0] - 2024-01-29
Added
- ResourceDefinition/VolumeDefinition to prometheus /metrics
- Added PeerSlots to ResourceGroups (create, modify and spawn)
- Added additional oversubscription calculation
- Added new oversubscription ratios to QSI result
- Prometheus/metrics added storage_pool to volume labels
- Added BalanceResourcesTask that in the first version tries to keep the resource-group place-count for resources
- Added MathUtils methods for integer-based square root, exponentiation, prime factorization, LCM and GCD calculation
- Added new checkExtFiles-API
- Schedule enable and snapshot restore now have a --storpool-rename option
- /metrics/scrape-target endpoint with drbd-reactor scrape targets
Changed
- Added copy of GenDbTables to GenCrdV* classes for migrations
- Escape executed commands in logs, making them paste-able
- Oversubscription now takes minimum of old and of new calculation
- Autoplace API answer now contains object refs to nodes and storagepools
- Make-Available will now retry auto-place with no restrictions on diskless resources
- StoragePoolMixing is now based on extent-sizes and thin/thick. Also changed implementation
- Autoplacer now also allows for mixed storage pools if the corresponding property allows
- Backup ship now uses the --target-storpool as a default storpool name
- ZFS and ZFS_THIN provider now report REFER for thin-volumes and snapshots and VOLSIZE for thick-volumes
- Remember node-ids for a later "forget-peer" when SkipDisk is enabled
Fixed
- Do not allow setting resource-groups with place-count 0
- Small DB resource leak in health check
- Fixed unrecoverable state with failed rollback of deleting snapshot
- Fixed race condition between node lost and backup queue cleanup
- NetCom SSL reimplementation: Fixed buffer handling, SSL handshake/negotiation/renegotiation
- Prohibit disk accesses in DrbdLayer while SkipDisk is set
Deprecated
- Support for EXOS
Removed
- Support for OpenFlex
v1.26.0-rc.1
[1.26.0-rc.1] - 2024-01-22
Added
- ResourceDefinition/VolumeDefinition to prometheus /metrics
- Added PeerSlots to ResourceGroups (create, modify and spawn)
- Added additional oversubscription calculation
- Added new oversubscription ratios to QSI result
- Prometheus/metrics added storage_pool to volume labels
- Added BalanceResourcesTask that in the first version tries to keep the resource-group place-count for resources
- Added MathUtils methods for integer-based square root, exponentiation, prime factorization, LCM and GCD calculation
- Added new checkExtFiles-API
- Schedule enable and snapshot restore now have a --storpool-rename option
- /metrics/scrape-target endpoint with drbd-reactor scrape targets
Changed
- Added copy of GenDbTables to GenCrdV* classes for migrations
- Escape executed commands in logs, making them paste-able
- Oversubscription now takes minimum of old and of new calculation
- Autoplace API answer now contains object refs to nodes and storagepools
- Make-Available will now retry auto-place with no restrictions on diskless resources
- StoragePoolMixing is now based on extent-sizes and thin/thick. Also changed implementation
- Autoplacer now also allows for mixed storage pools if the corresponding property allows
- Backup ship now uses the --target-storpool as a default storpool name
- ZFS and ZFS_THIN provider now report REFER for thin-volumes and snapshots and VOLSIZE for thick-volumes
- Remember node-ids for a later "forget-peer" when SkipDisk is enabled
Fixed
- Do not allow setting resource-groups with place-count 0
- Small DB resource leak in health check
- Fixed unrecoverable state with failed rollback of deleting snapshot
- Fixed race condition between node lost and backup queue cleanup
- NetCom SSL reimplementation: Fixed buffer handling, SSL handshake/negotiation/renegotiation
- Prohibit disk accesses in DrbdLayer while SkipDisk is set
v1.25.1
[1.25.1] - 2023-11-20
Fixed
- Fixed version mismatch check between new satellites and a pre v1.25.0 controller
- Fixed "node restore" to try to (re-) create tiebreaker resources if needed
- Fixed "node evacuate" no longer keeps diskless resource as a tiebreaker (in evacuating, but never deleting state)
- Fixed "sp l" showing spurious warning that a diskless SP is offline (due to caching bug)
- Fixed potential null pointer exception in AutoDiskfulTask
- Fixed "r c --drbd-diskless" falsely requiring free peer-slot
- Fixed bug where FILE/-THIN provider did not properly find existing snapshots (thus also not deleting them properly)