Release Notes

Synchronous replication

• Fixed an issue with synchronization/heartbeat channels unable to restore the connection when performance degradation occurs on the underline storage and VAAI is enabled for StarWind HA Devices.

Synchronous replication

• Fixed deadlock when a partner device is removed during full or fast synchronization.
• Fixed crash when updating StarWind service with mutual CHAP configured on StarWind HA devices.
• Fixed failure when extending StarWind HA device with disk-based synchronization journal.
• Fixed deadlock when a partner device is added.
• Fixed false fast synchronization when the first partner is added with a node majority failover policy.

Synchronous replication

• Fixed spurious replication failure.

Synchronous Replication

• Fixed the slow full synchronization issue.

Core

• Excluded asynchronous replicas, LSFS and LWC from the project.
• Encrypted secret passwords for chap authentications.
• Fixed the crash during the creation of the replica.
• Fixed the crash for time-outed I/O requests.
• Fixed the spurious lost connection.
• Updated the TLS version to 1.2 (used for email notifications).
• Minor fixes and updates.

Synchronous Replication

• Added synchronization journal places on a disk with two working strategies:
  
  Best Performance: The strategy provides better I/O performance while all device nodes are in a healthy state. But it can't guarantee fast synchronization in all cases.
  
  Fast Recovery: The strategy provides faster recovery in all cases but requires an I/O overhead for each write operation.
• Fixed the node majority logic during storage degradation.

Core

• Fixed the memory leak in the VSA-based version that could be caused by the clocks changing.

Core

• Fixed the crash for client write request handling.
• Fixed the deadlock for the processing of SCSI persistent reservation during client session registration.
• Fixed the issue with the service getting stuck when the log file would be located on the storage that had degrading performance.

StarWindX PowerShell Module

• Fixed memory leak on getting information for HA devices.

Core

• Fixed the issue with hang-up on service stopping. On service stop, it could get stuck on the device closing in configurations with parameter "QueueWorkersCount" value more than 1.
• Fixed the issue with hang-up on service stopping. In some cases, the service could not close the client connection gracefully. Now, the service closes it forcibly.
• Fixed the issue with hang-up operations on the underlying storage. Some RAID could produce hang-up operations that would never complete. Now, the service completes such operations with timeout as a workaround. The timeout default value is 60 seconds.
• Fixed the crash for client session closing.
• Fixed the issue when email notification could cause HA device timeouts and failure of the network connection between partners.
• Fixed the issue for STARTTLS authentication on SMTP — compatibility issues with some servers.
• Performance optimization for the VSA-based version.

Management Console

• Fixed control connection issues and improved logging and diagnostics abilities.
• Fixed the issue with service events flooding.
• Fixed minor issues in device creation wizards and dialogs.

StarWindX PowerShell Module

• Fixed the issue with size conversion during RAM device creation.

Synchronous Replication

• Fixed the issue with HA node synchronization rejection after service restart. When the service was restarted on one node, the synchronization procedure was not starting in some cases.
• Fixed the crash in the case when underlying storage or network connection between partners experienced performance degradation.

VTL and Cloud Replication

• Fixed the crash that happened during tape export in the case when cloud replication was configured with the option "Create new empty tapes automatically when existing tape removed from VTL for replication." When the option was enabled, the program could crash after the creation of one of the new virtual tapes.

Core

• Performance optimization for the VSA-based version.
• Fixed the issue with processing the close session operation in the state when system resources are extremely low.
• Fix for processing of the control connection close operation. In some cases, this error could lead to massive notification generation and make the management console unresponsive.
• Changed the session close procedure on service stopping. In case if a session cannot be closed gracefully, the service closes it forcibly. The service could hang up waiting for session closing in previous versions.
• Fixed the issue with Management Console hang-up when a free license is applied on the server without Internet access.

Synchronous Replication

• Fix for Node Majority failover strategy implementation. The failover routine was working incorrectly if the network connection between partner nodes was broken only in one direction.
• Fix for the auto restore procedure for three-way synchronous replication. Previous versions could be restored using the state information from two nodes of three. This is incorrect for the auto-restore procedure, as state information from the third node may affect the result of the restore procedure, too.
• Fixed the issue with losing the partner connection during the extend size operation. On some types of storage, this operation could take a long time. The waiting procedure has been changed to not cause partner disconnection.
• Fixed the error which led to full synchronization instead of fast synchronization in some cases.
• Fixed the error on processing IO requests that have stuck on underlying storage on one node in a device with three-way synchronous replication. This error led to the state when all three nodes worked incorrectly when the underlying storage on one of the nodes would not respond in time.

VTL and Cloud Replication

• MS Azure support: Added the option to upload virtual files to Archive tier directly without uploading to Hot/Cool tier and then moving it to Archive tier in the cloud.
• Added Write Protect option for virtual tapes.
• Fixed the error on restoring the tape split into parts from a cloud. When a tape was split into parts, some parts could not be downloaded from a cloud during the restore procedure.
• VTL file operation mode changed to close unused virtual tape files by default. If there are too many tape parts in use, keeping files open can lead to system resource exhaustion. This mode can be changed back to keep files open with the closedatafiles option in StarWind.cfg file.
• Auto-create option improved. Now it uses size and block size parameters of the previous tape to create the new tape automatically.
• The minimum tape size is limited to 50MB.

StarWindX PowerShell Module

• Physical devices are reported in the device list. It allows using PowerShell for exporting physical tape devices.
• MaintenanceMode property added to HA Device.

Licensing

• The License Agreement has been updated.

Core

• Default value for the iScsiPingCmdSendCmdTimeoutInSec setting has been updated to 5 seconds for software update on existing StarWind Virtual SAN installation.
• Fixed logging for email notifications: previously, messages with quotes from the mail server could be added to the internal log message and break the string format.
• Improved the service stop operation. In some cases, a client may keep the connection open even after the logical iSCSI connection shutdown. This prevents the normal stopping of the service. Now, Server closes the connection in these cases forcefully.
• Introduced a set of fixes and updates to the SCSI protocol implementation. Fixed the reporting of the size of MODE_SENSE length field in the response. The MODE_SENSE10 command is reported as unsupported by flat devices. Flat devices report the Resource Provisioned type instead of Thin Provisioned type on the Logical Block Provisioning VPD page. Resource Provisioned type suits better, as flat devices allocate all blocks at the start of the operation. UNMAP/TRIM commands are processed correctly as before and release blocks of underlying storage devices.

Synchronous Replication

• Added the option to use the SMB share as a witness resource for devices with synchronous replication and node majority failover strategy. Check StarWind vSAN online help for details (/help/).
• Processing of persistent reservations is fixed for devices with synchronous replication. At the moment, when the reservation state is being updated on one node, another node can report old reservation values to the client.
• Fix for updating the partner state in cases when the device was restored after a split-brain issue.

Management Console

• Fixed the crash in a case when StarWind Event log contains records with incorrect string format.
• Updates for the email notification settings dialog. Added validation for the list of email addresses in the "Send to" field. Login/Password fields are hidden in the case when Authentication is set to none.

StarWindX PowerShell Module

• Added functionality to create an HA device with Node Majority configuration. StarWind node or SMB Share can be used as a witness resource. See sample scripts CreateHAPartnerWitness.ps1 and CreateHASmbWitness.ps1.
• Fixed the processing of the ALUA parameter for an HA device creation commandlets.

VTL and Cloud Replication

• Cloud Replicator, Backblaze: updated error processing for the Backblaze HTTP client.
• Cloud Replicator, MS Azure: fixed performance statistics records in the cloud replicator log for MS Azure replication.
• Cloud Replicator, Amazon S3: China (Ninghia) region removed from the list.

Core

• Fixed the processing of the EXTENDED COPY(LID4) command. It was returning an error to the client on the previous build, thus making the client use non-VAAI commands.

Notifications

• Text file notification changed to open and close text file on each message instead of keeping the file open constantly.

E-mail Notifications

• Introduced a minor fix for secured authentication on SMTP — compatibility issues with some servers.

Synchronous Replication

• Full synchronization with CRC check, which has been added in the previous release, is now used only for cases when underlying storage is MS Storage Spaces. Synchronization type (Full or Full hash) is shown in the Management Console.
• For generic storage types old algorithm with data copy without additional CRC check is used due to better performance.
• Memory alignment fixed for data CRC request. Compatibility fix for VSA build.
• Fixed bug for the Node Majority configuration with two nodes and Witness node. Data corruption may occur in case when failure is caused only by the loss of the sync channel, and the failing node has the ability to process client requests at the moment when a loss of sync channel is detected.

Installer

• The number of files for log rotate is updated to the new value of 20 during the update of existing installations.

Flat Storage

• Improved the processing of responses to UNMAP/TRIM commands from the underlying storage system for ReFS: functionality to suppress log error on failed Trim call for the error 312 (ERROR_NO_RANGES_PROCESSED) added.

Management Console

• Fixed error messages from Event Log display in the notification area. In some cases, notifications have not been displayed when the option is configured.

StarWindX PowerShell Module

• Added sector size check for new Flat Image: it is not allowed to create Flat Device with a sector size of 512 bytes on disk with a physical sector size of 4096 bytes.
• Added parameter value validation for parameter "format" in Add-RamDevice call. Allowed values are fat16, fat32, ntfs, raw.
• Added "force" parameter for "remove" command.
• Added option to use custom barcode for new VTL Tape. See AddVirtualTapeCustomBarcode.ps1 sample.
• Added property CreateTapeOnExport for the VTL ApplyReplicationSettings() call. See VTLReplicationSettings.ps1 sample.

E-mail Notifications

• Fixed bug: sending E-mail Notifications using the SMTP server without authentication has been broken in the previous build.

Core

• StarWind Log Collector wizard added. Logs pack for support requests can be collected on the server using the Management Console. Use the corresponding context menu item for the Server to prepare an archive with logs and download it to the machine where the Management Console is running.
• Fixed the error when the device header file was referenced from the config file but did not exist on the disk, service could crash on start.
• Improved output of event log records for time ranges when the system time underwent changes.
• Improved the processing of responses to UNMAP/TRIM commands from the underlying storage system. In case when UNMAP/TRIM operation returned an unexpected result code, disk error was not reported to the client. In the previous versions, disk error was reported to the client, which showed up as a wrong "Bad block" error record in the client's logs.
• Added implementation for processing of Header and Data iSCSI digests on CPUs without the SSE4.2 command set.
• The default number of log files in log rotation settings increased from 5 to 20.

Synchronous Replication

• Improvement: full synchronization procedure is reworked to omit unneeded data movement. Instead of executing the full copy, it compares the CRC for data on partner devices block-by-block and copies the data only if blocks on HA partners are different.
• Fixed error: the server could crash when the partner node goes offline unexpectedly with high load present on the HA device.
• Default value for "ping partner" operations (iScsiPingCmdSendCmdTimeoutInSec) increased from 1 second to 3 seconds.
• Fixed the memory leak for the state when one node has RDMA configured and tries using the iSER for partner connection and the other node has no RDMA and doesn't accept the iSER connections.

E-mail Notifications

• Implemented the support for SSL/STARTTLS SMTP connections.
• Added the option to use SMTP Authentication.
• Added the option to test the current SMTP settings with the "Send Test Message" command in the "Modify Event Notification Rule" dialog for the Send E-Mail action type.

VTL and Cloud Replication

• List of regions for all replicators moved to an external JSON file and can be easily updated in case of changes on the side of the provider.
• Fixes for processing barcodes with non-default lengths (more or less than eight characters).
• Fixed the crash on attempt to upload the tape split into a large number of parts (order of thousands).
• Fixes for replicator settings management. Some controls in CloudReplicator Settings dialog were tied to incorrect parameters.

StarWindX PowerShell Module

• AddVirtualTape.ps1 was updated to use case-insensitive string comparison for a device name search.
• The “Remove-Device” command extended to use optional parameters for force disconnect, keep service data, remove stored data on device remove operation.
• The “IDevice::EnableCache” method added to enable/disable cache. PowerShell support added too. See FlashCache.ps1, FlushCacheAll.ps1 samples.
• Added the option to create HA devices with failover strategy. Use parameter $failover=0/1, see CreateHA*.ps1 samples.
• VTLReplicationSettings.ps1 — sample changed to use numeric values for replication destination types.
• "Valid" property added to the HA channel.
• MaintenanceMode.ps1 — better processing for boolean parameters when running from the command line.
• LSFS device type — AutoDefrag parameter removed, always true.

Management Console

• Minor Fixes and Updates.

Core

• Fixed the error in license validation: for licenses with a limited storage capacity of replicated devices, single node devices were taken into account by error.

Synchronous Replication

• Fixed the error with processing the "force remove" operation on the HA device. It could lead to service crashes in certain conditions.

VTL and Cloud Replication

• Fixed the issue with barcode generation: when new tapes were created in several tape drives of the same VTL device at the same time, the same barcode could be assigned to different tapes.

Core

• Added support for iSCSI Header Digest and iSCSI Data Digest. These features provide CRC data control in iSCSI headers and PDUs, respectively, and can be enabled by setting corresponding options in iSCSI Initiator on the client.
• Fixed memory leak in Flash Cache during multiple device add/delete operations.
• Optimized performance when more than three sessions connect to the target. SMI-S management for HA has been fixed to report HA devices correctly.
• Added log message for error on session reinstantiation due to the stuck old session.
• Fixed issue in cache deinitialization. In some cases, devices with disk cache could prevent normal service termination.

VTL and Cloud Replication

• Added option to cancel current cloud upload or cloud download operation.
• Extended replication functionality. Now in the replication process, several objects can be processed in parallel to remove the performance limitation of a single data object (for MS Azure, please see "Target throughput for single blob" at the following link: https://docs.microsoft.com/en-us/azure/azure-subscription-service-limits).
• Wasabi replicator: EU-central-1 region added.
• Added functionality to manage Cloud Replicator settings using UI (Server - Configuration - Cloud Replicator Settings).

Synchronous Replication

• Added option to change ALUA settings of an existing device.
• Optimized processing of failover to minimize response time when a partner server has gone offline without proper shutdown. Please, see KB for more information.
• Fixed processing of sync channel loss. Failover procedure involved the request of some extended status fields from the partner nodes, which could take long and exceed command processing timeout. This state could occur in some cases when running Windows Server 2019 and could lead to data loss.
• Changed failover procedure to process sync channel loss without extended interaction with partners.
• Optimized performance for NUMA systems. In the previous build, there was an issue with correct NUMA-node binding.

Management Console

• Optimized UI performance.

StarWindX PowerShell Module

• Added option to set parameters for most of the sample scripts.

Log/Write Cache (experimental)

• Added auto-start for snapshot migration procedure. In previous builds, snapshots have been kept on the log without moving to the base device, filling up the disk where LWC log resides.

LSFS

• Fixes for issues related to device start after incorrect server shutdown.
• Fixes for device operation with updates from the previous versions.

NVMe-of Target Demo

• Demo target updated, fix for processing Max Queue Size option added.

Core

• Fixed SMI-S management issues. Now, it works correctly with single-node and multi-node devices.
• Fixed deadlock of the service management module. Occasionally, this issue occurred after UI requested the device information while the service core was creating a device with L2 cache. That deadlock blocked any subsequent operations, like device creation/deletion, session registration, etc.
• Fixed incorrect processing of session disconnections initiated by iSCSI Initiator during the iSCSI Login stage. Previously, these issues could cause the service to crash.

VTL and Cloud Replication

• Added support for Azure Government storage in the replicator.
• Added option to schedule replicated virtual tape file deletion from public cloud storage.
• Added option to delete virtual tape files from the local and public cloud storage via GUI.
• Added support for Signature Version 2 for S3-compatible Cloud Storage replicators.
• Fixed replication process freezes happening after Backblaze replicator received an unexpected message from the server. The replication progress got stuck, and the status of the tape remained unchanged. The log was flooded with "Inner exception message: Exception of type 'System.Net.Http.HttpRequestException' was thrown" records.
• Fixed error with updating the list of available replicators observed after installation of a new build over a previous version. In build 12658, after updating from the previous versions, S3-compatible cloud storage and Wasabi Cloud Storage options were unavailable in replication settings.

Synchronous Replication

• Fixed client connection breaks after switching to the Maintenance mode and synchronization status change. When client sessions were executing long-time operations, the client could be disconnected if the device status changed.
• Fixed the NVMf Target Service update procedure: during installation of build 12585 over the existing installation of the same version, NVMf Target was not reinstalled correctly. Now, NVMf Target config is reset by the installer to avoid any possible issues due to NVMf Target misconfiguration.

Synchronous Replication

• Fixed the issue with incorrect processing of storage hardware failures (introduced in build 12393). Before, in build 12393, disk storage failures on an HA node could sometimes be left unprocessed. After the hardware issue was resolved and the HA node restarted, no correct full sync process was done, leading to the risk of data corruption.
• Updating is strongly recommended for systems that are running V8 build 12393 and later versions!

VTL and Cloud Replication

• Added support for Wasabi Cloud Storage in the replicator.
• Added support for S3-compatible Cloud Storage in the replicator. Signature version 4 is supported now, Signature version 2 support will be added in the upcoming update.
• Fixed restore functionality for the replica in MS Azure Cloud. Before, if data had resided in the archive when restoring, the restore process would have never ended.

NVMf Target Demo

• Fixed the connection error for Management Console occurring when the NVMf Target Demo service was not installed on a server.
• Fixed the NVMf Target Service update procedure: during installation of build 12585 over the existing installation of the same version, NVMf Target was not reinstalled correctly. Now, NVMf Target config is reset by the installer to avoid any possible issues due to NVMf Target misconfiguration.

VTL and Cloud Replication

• Added support for LTO8 tape devices.
• Added new PowerShell commands for managing VTL devices and Cloud Replication settings.

See the samples for more information.

NVMf Target Demo

• Simple NVMf Target can be created for demonstration purposes. The existing NVMf initiator can be used now to connect the target.
• Simple RAM Disk and Flat Storage devices are supported. Mellanox RDMA-enabled network adapters with the latest drivers from the manufacturer are recommended.
• Other adapters that provide the NetworkDirect API layer can be used.
• Please take into account that StarWind vSAN installation process overwrites NVMf Target configuration file (nvmf.conf). If an old nvmf.conf file exists before installation, it will be saved as nvmf.conf.bak.
• Unless the used StarWind Virtual SAN version supports the feature, download the new one and request the new license key. This works for Free and Trial licenses too.

Check our Web site for more information.

Synchronous Replication

• Fixed 100% CPU load issue. Before, when HA device started working under irregular workloads, sync channel transport would load one CPU core by 100% in some cases. If the number of sync channels is the same as the number of server CPU cores, it could make the service unresponsive.

Management Console

• Help moved. Find it by the following link: https://www.starwindsoftware.com/help/
• Other useful links were added to the Help menu.

Core

• Fixed service start crashes in systems where NVMe devices are present.

VTL and Cloud Replication

• Added support for MS Azure Cloud Storage in replicator.
• Fix for performance issue that was present in 12146, 12166 builds: VTL write performance was affected by non-buffered file operation mode.
• Implemented support for offline pre-seed scenario. If tape backup has been added to cloud storage using offline data transfer service or other methods, tapes can be registered on offline shelf to make them available for restore commands. Use “Restore from Cloud” dialog, select tapes and check “Add tape record from cloud to shelf without downloading it” option.

Core

• Notifications: Event ID code added to message string for notifications to text file.
• Fix for processing unfinished iSCSI Tasks: when iSCSI session terminates in non-graceful way, last commands could remain unprocessed, preventing session termination. These sessions prevent the service from correct shutdown with log message like “Waiting for 5 seconds for clean up to complete (XX refs)..”.
• Fix for operating with ProxMox Virtual Environment. Fixed processing of WRITE_SAME16 SCSI command. StarWind Service was crashing on virtual disk initialization when used with ProxMox VE and VAAI commands enabled.

Synchronous Replication

• Fix for PowerShell command to enable setting two or more synchronization or heartbeat channels for HA device.
• Fix for processing configuration change when network interface is being removed from the node and then added back immediately using Replication Node Interfaces wizard. When HA device is configured in Node Majority mode, this issue could lead to data corruption.
• Fix for processing configuration change when non-active HA node is being removed from 3-node HA device configuration with some client load present. When HA device is configured in Heartbeat mode, this issue could lead to loss of one of two remaining nodes: service becomes unresponsive.

Log Write Cache

• Log Write Cache functionality added as experimental feature. Log cache works as L2 write-back cache with data logging. This helps increase write speed in some workloads. In scenario with Synchronous Replication LWC helps processing incorrect node shutdown:
  - eliminates need for full synchronization after incorrect node shutdown;
  - allows auto-sync with correct HA partner after full blackout state.
• Use Add Device Advanced wizard, "Thick-provisioned with Write Log" option to create flat device with write log.
  
  Check our Web site for more information.

This functionality is not for production use yet.

Flat Storage

• Fixed issue with the processing of Trim/Unmap commands. This error has appeared in the previous build. Symptoms of the error are "Bad Block" error messages in Windows System Event log on a client system and messages in StarWind log files at the corresponding time: IMG: *** ImageFile_FileLevelTrim: FileLevelTrim: FSCTL_FILE_LEVEL_TRIM failed (err 6). These messages indicate the error in the processing of Trim/Unmap command on a virtual device. No data is lost and no disk errors are present.

LSFS

• Fixed issue on pre-V8R6 devices. When LSFS-based 3-node Synchronous Replication that has been created in V8R5 version (build 11456 or earlier versions), upgraded to new build, synchronization could fail.

Management Console

• Added support for upcoming StarWind VSA 2.0.

VTL and Cloud Replication

• Support for Backblaze B2 Cloud Storage has been added in the replicator.

LSFS

• Fixed issue on pre-V8R6 devices: when user data size exceeded 1.2 TB, data corruption could occur on the service restart.
• Fix for excessive storage usage under high load: defragmentation procedure has been changed.

Synchronous Replication

Event log messages became more explicit to identify the source of performance issues:

• (id 809) Underlying device response time is longer than expected.
• (id 824) Partner "%s" request response time is longer than expected.
• (id 825) Request execution time is longer than expected.

The maintenance mode message has been changed in accordance with “on” and “off” event:

• (id 821) Maintenance mode is turned ON.
• (id 822) Maintenance mode is turned OFF.

• Fix for processing HA device configuration changes with Node Majority strategy. When two HA partner devices were deleted manually on a server, the remaining partner (without the majority) was staying online and processing requests. Now, it is turned to the "not synchronized" state, as it has not enough votes for the majority.
• Fix for operation with Node Majority strategy with a Witness node. When one data node was offline and the Witness node went offline and then back online, the HA device did not return to the online state even having the majority with nodes.

E-mail Notifications

• The 30-second interval between sending separate messages has been removed.
• Fix for an issue when SMTP port number couldn't be changed.

Flat Storage

Improvement: when creating a device on storage that reports physical sector size incorrectly, the service allows creating the device with a warning.

StarWindX PowerShell Module

• The VTL management functionality has been added to StarWindX PowerShell Module. You can find the samples below:
• \StarWindX\Samples\powershell\CreateVirtualTapeLibrary.ps1
• \StarWindX\Samples\powershell\InsertVirtualTape.ps1
• \StarWindX\Samples\powershell\RemoveVirtualTape.ps1
• Powershell module: Fix for reporting tape barcode in device.Slots object.

A sample for a 3-node LSFS-based HA device has been added:

\StarWindX\Samples\powershell\CreateHA_3_LSFS.ps1

• Fix for processing "SectorSize" parameter. Now it is possible to create devices with a 4k sector size.
• Fix for the HA device creation with COM object and PowerShell scripts.
• Fix for the snapshot creation functionality in COM object and PowerShell scripts.

Known Issues:

• Issues in the upgrade of LSFS-based 3-node Synchronous Replication that has been created in V8R5 version (build 11456 or earlier versions). The fix is planned for the next update. The device can be switched to a two-node mode or to a flat-storage based mode as a workaround.
• Issues in VTL devices operation with MS DPM 2016. MS DPM 2013 works without issues. This case is under investigation now.

Synchronous Replication

• The node majority failover strategy has been implemented in addition to the existing heartbeat mode. It can be used when the second physical channel cannot be provided for a heartbeat between the nodes. An odd number of nodes is required for operation in this scenario. A witness node without storage can be used in the case when the third storage node with its own copy of data is excessive. Please check the online documentation for more information.
• The maintenance mode for devices with Synchronous Replication has been added. When a device is set to the maintenance mode, the device nodes can be switched on/off freely, and the full synchronization is not required when the device is returned to the operational mode. Clients must be disconnected before switching the device to the maintenance mode.

LSFS

• The LSFS engine has been improved to get better performance, faster initialization time after a graceful shutdown and after an incorrect shutdown. The ability to keep metadata on SSD storage instead of RAM has been added. It is recommended to re-create LSFS devices and migrate data to new devices.

Cache

• Cache performance has been improved in both Write-back and Write-through modes.

E-mail Notifications

• The timestamp has been added to the message text in an e-mail.

VTL and Cloud Replication

• The support for AWS GovCloud (US) Region in the replicator has been added.
• The processing of timeouts and connectivity issues on operations with Cloud Storage has been improved.

Core

• General performance improvements and fixes.
• The processing of task queues on high load and on connection loss has been improved.

Core

• Fixed issue with hang-up on service stopping. On service stop it could get stuck with repeating messages in log: "Waiting for 5 seconds for cleanup to complete (XX refs)...". Messages are repeated without changing the number of references and without actual stop of execution.

Synchronous Replication

• Fixed issue with incorrect processing of time-outs for commands on synchronization channel. Error in processing of timed-out commands could lead to crash of the service.
• ODX performance optimizations.

Cache

• Fixed processing of TRIM/UNMAP command. When cache memory allocation was still in progress, but virtual devices have already been set to the functional state, service could crash on receiving TRIM/UNMAP command.

Synchronous Replication

• Fixed issue with "Synchronization request will be refused" state in logs. In some cases, devices with Synchronous Replication could not start synchronization after a server reboot, and the message "Synchronization request will be refused" appeared in service logs.

Known issue

• LSFS-based devices with Synchronous Replication could lose synchronization when disk load was growing significantly due to replica synchronization or cache flushing.