Breaking changes

• [Backups] compression algorithm was changes from gzip/deflate to zstd, which might introduce some backward compatibility concerns. Please read the 'Backward compatibility' section in this article to get more information.

Server

• [Backups] switched to zstd compression algorithm for all backup types and exports. More info here
• [Backups] added 'Direct Upload' mode. More info here
• [Cluster] fixed issue with single-node cluster setting it's commands version to 0
• [Configuration] added configuration option 'Backup.Compression.Algorithm' was introduced (Zstd by default)
• [Configuration] added configuration option 'Backup.Compression.Level' was introduced (Fastest by default)
• [Configuration] added configuration option 'Backup.Snapshot.Compression.Algorithm' was introduced (Zstd by default)
• [Configuration] added configuration option 'Backup.Snapshot.Compression.Level' was introduced (Fastest by default)
• [Configuration] added configuration option 'Export.Compression.Algorithm' was introduced (Zstd by default)
• [Configuration] added configuration option 'Export.Compression.Level' was introduced (Fastest by default)
• [ETL] fixed error handling and TransactionalId generation for Kafka ETL
• [ETL] fixed issue with ETL not entering fallback mode when cluster is unstable, resulting in sending the same batch again
• [ETL] fixed secured channel detection for RabbitMQ and Kafka ETLs
• [HTTP] switched inter-server communication from HTTP/1.1 to HTTP/2 protocol
• [HTTPS] switched to 'SslStreamCertificateContext' to improve SSL authentication performance on Linux
• [Indexes] added support for equality comparers for floats and decimals in static indexes
• [Licensing] fixed issue with force update of expired license not returning unsuccessful state to the user
• [Licensing] properly detect revision compression when switching a license to Community
• [Replication] validate if certificate has write permissions
• [Replication] fixed issue with attachment streams existence for documents when conflicts are being solved manually
• [Replication] fixed issue with deletion of non-replicated tombstones
• [Replication] fixed issue with missing revision tombstones, because revision polices are being applied
• [Subscriptions] fixed possible NRE when processing concurrent subscriptions with criteria
• [Voron] fixed possible AVE when doing background flushes
• [Voron] fixed AVE during compaction on x86 or when we surpass MaxScratchBufferSize
• [.NET] updated to .NET 7.0.13

Client API

• [Compatibility] fixed FHIR compatibility
• [HTTP] switched Gzip compression mode to 'Fastest' from 'Optimal'
• [Querying] fixed issue with missing parenthesis when combining 'Where' and 'Any' clauses when querying using LINQ
• [Querying] fixed query hash calculation for empty string parameters
• [Session] added 'WhatChangedFor'

Studio

• [Querying] added the ability to set Projection Behavior when querying
• [Kafka] fixed issue with setting Kafka Connection String with plain certificate

Server

• [ETL] fixed change detection in Queue ETL
• [ETL] better handling of cancellation tokens during 'InitTransactions' call
• [ETL] disposing producer on a failed 'InitTransactions' call
• [Indexes] fixed issue with enabling auto indexes after they were disabled cluster wide
• [Revisions] fixed possible NRE when enforcing revision configuration on tombstones
• [Subscriptions] memory optimizations while processing subscription documents
• [.NET] updated to .NET 7.0.11

Studio

• [Migrations] allow 'srv' in mongo connection string

Server

• [Backup] fixed issue with aborting One-Time backup not killing the operation
• [Backup] fixed edge case where backup would not failover properly if no successful backups are done
• [Backup] fixed edge case where backup status would not be updated if database is idle
• [Backup] blocked deletion of database if restore is in progress
• [Configuration] added 'Http.Http2.MaxStreamsPerConnection' and set it to unlimited value
• [Cluster] fixed issue with HIGH CPU when there is a single node cluster
• [Cluster] dedicated thread for cluster transaction execution
• [Documents] fixed issue with huge documents not showing the ID of the document in some case
• [Indexes] added support for 'default(DateTime)'
• [Linux] detecting running close to OS limits (notification and SNMP endpoint)
• [Querying] projections will return '@id' property for simple projections with loaded documents returned
• [Querying] decreased memory usage when streaming compressed documents
• [Replication] decrease memory used when large number of compressed documents/revisions is skipped
• [Revisions] removed the need to decompress the revision when training a compression dictionary
• [Revisions] fixed various cases when revisions are being created or purged
• [Revisions] added ability to enforce configuration for specified collections only
• [Revisions] changed default number of conflicted revisions from unlimited to 1024 and added a notification if we exceed that number
• [Subscriptions] decreased the cluster pressure for ACK commands
• [Subscriptions] fixed issue with removing database would not clear all of the subscription states
• [.NET] updated to .NET 7.0.10

Client API

• [Bulk Insert] will send heartbeats to keep operation alive (requires server update as well)
• [HTTP] properly taking into account if node is in Rehab when failover is happening
• [Operations] fixed issue with 'WaitForCompletion' hanging when operation is aborted but additional task is not
• [Session] fixed issue with metadata properties starting with '@' got cleared after 'SaveChanges'
• [Session] fetch atomic guard of a missing document in cluster transaction to avoid desync of atomic guards (requires server update)
• [Querying] allow Proximity search 0

Studio

• [Backup] removed the ability to delay one-time backup
• [Backup] added copy button to restore log

Server

• [Attachments] when 'WaitForIndexesAfterSaveChanges' is used - attachments and their parent document collections are taken into account
• [Backups] support for 'Refresh' in Smuggler
• [Backups] Smuggler now skips Counters, Time Series, etc when parent document is being skipped via TransformScript
• [Backups] file name and counters are added to restore progress
• [Cluster] fixed issue with compare exchange tombstones not being cleaned when replication factor is less than number of nodes in the cluster
• [Cluster] sending early to the Client API that topology (Cluster and Database) needs to be updated
• [Configuration] setting 'Databases.MaxIdleTimeInSec' can now also be set per database
• [Connection Strings] validating factory name before connection string is added
• [Import] added ability to add reference key to embedded object
• [Indexes] optimized Bloom Filter performance and indexing time of the first indexing batch
• [Indexes] fixed issue with Counter indexes not indexing counter group when batch is being canceled due to low memory
• [Indexes] optimized memory (managed and unmanaged) allocations in Lucene
• [Indexes] added support for more expression types in Index Cleanup
• [Linux] fixed cgroup parsing issue on newer Ubuntu
• [Monitoring] added Prometheus integration
• [Operations] avoid using HttpContext.RequestAborted cancellation token for background operations - it is being reused by the ASP.NET and can cause early operation cancellation
• [Querying] fixed issue with projection function returning incorrect document ID and ability to project transitive dependencies
• [Querying] decreased the amount of cached loaded documents to 16k when projection is used
• [Querying] support for surrogate characters in alphanumerical sorting
• [Revisions] fixed issue with forced revision not marking a document with HasRevisions flag and replication of such documents
• [TCP] correctly detect using which address (internal or external) connection should be made
• [Tombstones] better notification when tombstones cleaner is getting stuck because of a disabled background task
• [Voron] fixed possible AVE in DecompressionBuffersPool.GetPage
• [.NET] updated to .NET 7.0.9

Client

• [HiLo] exposed 'GenerateDocumentIdAsync' in 'IHiLoIdGenerator'
• [HTTP] fixed possible NRE in NodeSelector when FastestNode mode is used
• [HTTP] avoid leaking HTTP request when speed test is being performed when FastestNode is used
• [Indexes] fixed issue with generation of index with nested dictionaries
• [Operations] added 'Kill' and 'KillAsync'
• [Session] ability to pass HighlightingOptions via DocumentQuery

Server

• [.NET] updated to .NET 7.0.8 to fix functional regression in X.509 Certificate imports introduced by MS in .NET 7.0.7 and described in KB5028608