What's new

Who should upgrade

Customers hosting Server on Windows 2012 R2 or older should disable HTTP/2 support by setting 'Http.Protocols' configuration option to 'Http1'.

Server

• [Configuration] added 'Http.Protocols' configuration options that allows to specify supported HTTP protocols by the server

Server

• [Backup] added progress when restoring a huge file to indicate that restore is actually taking place
• [Backup] number of concurrent backups is set by default to number of licenses cores / 2 or to Backup.MaxNumberOfConcurrentBackups value
• [Backup] reduce memory pressure when export is being done
• [Backup] overdue backups when switched to a different node should have 1 minute delay to wait if the original node will recover or not
• [Certificates] disabled downloading the CRL file
• [Certificates] fixed issue with replacting the certificate when it has a password
• [Configuration] added Indexing.ErrorIndexStartupBehavior
• [Configuration] added Memory.MaxContextSizeToKeepInMb
• [Configuration] added Cluster.TimeBeforeRotatingPreferredNodeInSec
• [Configuration] added Security.TlsCipherSuites
• [Configuration] added Security.Certificate.Validation.Exec, Security.Certificate.Validation.Exec.Arguments and Security.Certificate.Validation.Exec.TimeoutInSec
• [Cluster] fixed possible timeout when leader received same command twice
• [Cluster] when follower ambassador is being closed, set proper disconnected reason
• [Cluster] Compare Exchange tombstone cleanup should continue executing if there is more work to be done instead of waiting for next execution
• [Cluster] database should not be places as preffered note when it is still loading
• [Cluster] fixed possible deadlock in remote connection that could happen during server shutdown
• [Debug] added /admin/debug/dump
• [Debug] added /admin/debug/gcdump
• [Expiration] use same date among all batches during expired documents cleanup
• [Expiration] cleaner will correctly continue removing the expired documents if there is more to delete instead of waiting for next execution
• [Indexing] enhanced validation for pattern for reduce output documents
• [Indexing] better behavior when updating pattern for reduce output documents
• [Indexing] added the ability to specify pattern for reduce output documents collection name
• [Indexing] support for using multiple string interpolations in the PatternForOutputReduceToCollectionReferences expression
• [Indexing] fixed issue with JavaScript indexes not respecting Indexing.MaxStepsForScript configuration option in some cases
• [Indexing] proper field options inheritance from default field options in JavaScript indexes
• [Indexing] proper behavior when arrays are being indexed in JavaScript indexes
• [Indexing] properly handle out of disk space errors when this happens during Lucene write
• [Indexing] properly flatten the arrays when converting to Lucene
• [Indexing] avoid NRE in JavaScript index when converting output to BlittableJSON
• [Indexing] avoid possibility of infinite indexing batch loop when we have LoadDocument
• [HTTP] added support for HTTP/2
• [Licensing] updated EULA
• [Logging] moved behavior from per thread to per processor core
• [Major][Memory] Revamped memory management from per thread model to per processor core model
• [Memory] fixed a possibility that available memory would result in a negative number
• [Memory] replaced ThreadLocal with LightweightThreadLocal to reduce the GC pressure
• [Memory] decreased allocationg when using XXHash64Context and XXHash32Context
• [Memory] large contexts are being disposed right away instead of being kept in the pool
• [Memory] on low memory dispose contexts that were created in previous generations of the pool
• [Memory] on low memory release unused threads from the pool
• [Memory] on low memory clear indexed terms cache
• [Monitoring] when instance was never accessed SNMP endpoint for last non-cluster admin access time will return max value
• [Monitoring] time since oldest backup endpoint in SNMP will use the same algorithm for calculation as Databases View in Studio
• [PAL] added better detection for out of memory and out of disk space in Windows
• [Querying] fixed possible NRE when exporting to CSV query results from map-reduce index
• [Querying] correctly calculating index etag when reference collections are used
• [Querying] fixed possible NRE when doing a dynamic query and retrieved index would be removed after picking it
• [Querying] support for projecting from a document that contains $type and $values fields
• [Querying] support for projecting from a document that contains $type and $value fields
• [Querying] fixed NRE when projecting from Compare Exchange using JavaScript projections
• [Querying] do not apply query optimization when outer query has boost defined
• [Replication] fixed conflict status detection when we have unused database IDs
• [Revisions] putting revision tombstones only once to avoid possible endless replication loop
• [Revisions] prevent from creating a duplicated revision when incoming revision and the document are the same, but change vector contains unused database ID and resolved flag
• [Tombstones] cleaner should execute operations in batches to reduce memory pressure
• [Subscriptions] added the ability to use properties from @metadata that need to be escaped in the query
• [Security] added TLS 1.3 support
• [Major][.NET Core] updated to .NET Core 3.1.2
• fixed possible Etag overflow on change vector merge

Client API

• [Conventions] added HttpVersion convention
• [Indexing] added the ability to specify collection name manually in LoadDocument
• [Session] NoTracking session should not throw when document does not exist on the server
• [Operations] added DocumentStore.Maintenance.Server.ForNode method
• [Querying] always quoting collection names

Embedded

• added RestartServerAsync method
• added ServerProcessExited event

Studio

• [Documents] added the ability to copy change vector and document ID
• [Export] added missing configuration options
• [Indexing] fixed issue with saving an index with map when bracket access is being used in the function
• [Indexing] check for duplicate fields when saving an index
• [Patching] ability to define query options when patching
• [Security] properly disable sections of the Studio according to the security clearance
• various fixes and stability improvements

Server

• [Cluster] various stability improvements
• [Cluster] fixed possible socket leaks in FollowerAmbassador and CandidateAmbassador
• [Cluster] large snapshot installation will be done via file from now on to not hold the connection and avoid encountering read timeout
• [Backup] fixed possible memory leak during export that could cause decreased speed of export
• [Backup] fixed a possibility of creating duplicate backups
• [Backup] fixed possible failure when doing snapshot and incremental backup when encryption is set
• [Backup] optimized restoring from Google Storage process
• [Backup] added support for backup retention policy on Azure
• [Indexing] fixed updates of Reference documents when OutputReduceToCollection is set
• [Indexing] OutputReduceToCollection should not create new side-by-side instances after server restart
• [Indexing] processing speed will include side-by-side indexes and only when index is running
• [Indexing] separating MapReferenceSuccesses in statistics from MapSuccesses
• [Indexing] handling nested null propagation operators and null conditional access
• [Migration] added a support for case sensitive collations
• [Monitoring] added server certificate expiration endpoint to SNMP
• [Monitoring] added time since oldest backup endpoint to SNMP
• [Monitoring] added the ability to disable time window checks that are problematic for some monitoring software
• [Patching] exposing @change-vector in metadata
• [Replication] preventing double dispose of replication handler which could cause a replication to stop completely in some cases
• [Replication] fixed possibility for a deadlock of Replication by Tombstone Cleaner

Client API

• [Commands] added support for Broadcasting along with 'FirstBroadcastAttemptTimeout' and 'SecondBroadcastAttemptTimeout' conventions
• [Conventions] added 'HttpVersion' convention with HTTP/1.1 as a default
• [Operations] handling properly server down when waiting for an operation
• [Projections] support for IDictionary in JavaScript projections
• [Projections] added support from projecting from a dictionary when dictionary key has a comma inside
• [Projections] support for projecting null array
• [Querying] fixed issue with creating a query on Id field when OrderBy is used
• [Session] Load should not throw when 'NoTracking' is enabled and document does not exist on the server
• [Subscriptions] added ToSubscriptionCreationOptions that allows to convert from generic to non-generic subscription

Studio

• [Database] added the ability to modify database record
• [Documents] fixed issue with adding same attachment which would not trigger popup window
• [Indexing] allow to filter indexes that has errors

Tools

• [Recovery] better error handling when recovering encrypted storage without encryption key

Server

• [AuditLog] Avoid throwing ObjectDisposed during audit logging if the socket was closed
• [Backup] Fixed issue that an idle database wakes up but doesn't run backup
• [Indexing] Fixed error logging if there was an error in the referenced document
• [Indexing] Fixed saving of reduce documents to OutputToReduceCollection (System.FormatException: Bad variable size int)
• [Replication] Handled a conflict should take into account the unused databases
• [SNMP]Fixed database startup load due to exceptions thrown by SNMP watcher
• [Subscriptions] Handled passive nodes during subscription connection
• [Subscriptions] Improved mechanism to move between nodes in case we don't know what the responsible node should be
• [Subscriptions] Added timeout when there is compare exchange subscription substitution concurrency to prevent a partial live lock
• [Subscriptions] Fixed double dispose resulting in unexpected signaling of ConnectionInUse
• Fixed versions of dependency DLLs
• Updated to .NET Core 2.2.8

Client

• Added XMLdocs to the NuGet pckaged
• Disabled MaxIdleTime of ServicePoint

Server

• [Backup] Backup task will delete the .in-progress file when it fails
• [Backup] Avaliable space validation before the backup
• [Cluster] Fixed node kicked out of the cluster if the we have election timeout mismatch
• [Cluster] Added ability to specify topology priority order
• [Configuration] Increased default value of Databases.ConcurrentLoadTimeoutInSec setting from 10 to 60 seconds
• [Indexing] Added ability to define pattern for reference documents of OutputToCollection docs generated by map-reduce indexes
• [Indexing / Voron] Fixed unhandled case of leafs compression which resulted in map-reduce indexing error
• [JSON] Fixed memory overlap on ID slice creation, when we have two or more control characters in a row
• [Replication] Fixed replication looping when having faulty replication configuration with two replication channels from the same source to the same destination
• [Revisions] Allow to define separate conflict revision configuration
• [Stats] Exposed IO metrics for the system storage

Client

• [Index Definition] Fixed needless usage of ToDictionary() for simple calls
• [Index Definition] Fixed missing int cast on method call arguments
• [Operations] Synchronization of database name validation between Client API and server side
• [Operations] Added ConfigureRevisionsForConflictsOperation
• [Projections] Added support for nullable DateTime in projection
• [Projections] Fixed JsConverter strip 'Value' or 'Key' from dictionary entity
• [Projections] Fixed unhandled case of projection with SelectFields
• [Session] Fixed entities loading during OnBeforeStoreEvent
• [Session] Propertly converting dictionary stored in metadata
• [Session] Added active async tasks tracking and throwing on session dispose if there are any

Studio

• [Stats] Added IO metrics view for the system storage