What's new

Server

• [Backups] fixed issue with executing Backup Retention policies on Azure
• [Backups] allow to restore a document when collection was changed in the same incremental backup
• [Backups] fixed possible memory leak during import when a lot of compare exchange values are being imported
• [Backups] fixed issue with editing server-wide backup tasks that could cause those tasks to hang
• [Databases] before deleting a database a database path intersection is checked with other databases
• [Debug] fixed issue with capturing stacktraces for entire cluster
• [Debug] fixed start date of a cloned transaction in /admin/debug/txinfo endpoint
• [Databases] fixed NRE that could happen if database was created with Client Configuration set
• [Indexes] release temporary files on low memory to reduce the memory pressure further
• [Indexes] fixed a race between creating a new auto index and deleting the superseded one
• [Indexes] fixed usage of concurrent indexes semaphore to avoid NRE
• [Indexes] fixed issue with using ThenBy and ThenByDescending in indexes
• [Indexes] fixed issue with using Enumerable.Range in indexes
• [Indexes] added support for FindIndex and FindLastIndex methods in indexes
• [Indexes] when updating NuGet package in index to a newer version no longer restart of server is required
• [Indexes] support for native DLLs from NuGet
• [Indexes] fixed null string concatenation behavior in indexing - now it matched C# one
• [Querying] fixed possibility of sending an invalid JSON when returning facet results
• [Querying] fixed possibility of NRE when projecting same document multiple times during streaming
• [Querying] decreased memory usage when executing Distinct queries
• [Replication] fixed memory allocation when handling tombstones as a conflicts resolutions
• [Replication] fixed NRE on refresh of document that was added in cluster transaction
• [Voron] making sure that we will not remove currently used scratch buffer file from the list of active scratches during transaction rollback
• [Voron] making the acquirement and usage of a pager state more robust to avoid accessing already disposed pager which could result in AVE
• [Voron] preventing a racy access and dispose of a scratch file between the storage cleanup and creation of a new transaction
• [Voron] fixing the case that after the import of a database we didn't flush all journals if there was no activity on a database what resulted in large number of scratch files
• [Voron] using a correct pager state when prefetching segments
• [.NET] updated to .NET 5.0.6

Client API

• [Events] added OnSessionDisposing event
• [Events] fixed issue with OnBeforeQuery event not being invoked when streaming
• [JSON] internal enumerable converters need to take into account that a custom converter could have been supplied by the user
• [Session] fixed caching when Lazy is used
• [Session] properly incrementing number of requests when attachments and revisions are used
• [Session] added load revisions lazily implementation
• [Session] added AddOrPatch and AddOrIncrement methods
• [Subscriptions] fixed possibility of context leak when processing subscriptions and error has occurred
• [Smuggler] fixed possibility to context leaking during Import
• [Querying] added support for aggregation queries when raw query is used

Studio

• [Indexes] expose index wait time on performance graph when max number of concurrent indexes is set
• [Indexes] fixed issue with indexes getting unselected when progress is being updated
• [Indexes] fixed issue with displaying tooltips on performance graph
• [IO Stats] fixed issue with importing I/O stats from the debug package
• [Refresh] fixed issue with updating the refresh frequency
• [Traffic Watch] display also seconds in request time
• [Traffic Watch] added TCP connections

Server

• [Indexes] fixed issue where index replacement, therefore index update, could cause index to become paused
• [Indexes] fixed possible deadlock in indexing that could occur when 'Indexing.MaxNumberOfConcurrentlyRunningIndexes' is used
• [TCP] gently closing TCP TestConnection and TCP Ping to avoid moving into CLOSE_WAIT state

Server

• [Backup] Fixed the restore of tombstones from a backup
• [Cluster] Added endpoint to remove entry from Raft log
• [Configuration] Added Indexing.MaxNumberOfConcurrentlyRunningIndexes configuration option (default: null - no limit)
• [Configuration] Increased dafault value for Indexing.Encrypted.TransactionSizeLimitInMb configuration option depending on the amount of available memory:
  • <= 1GB of RAM or running in 32bits - 96MB
  • <= 4GB of RAM - 128MB
  • <= 16GB of RAM - 256MB
  • <= 64GB of RAM - 512MB
  • > 64GB of RAM - 1024MB
• [Configuration] Added Http.AllowSynchronousIO configuration option (default: false)
• [Indexing] Fixed the behavior of recreating indexing writer upon every commit
• [Indexing] Fixed handling for ERROR_COMMITMENT_LIMIT to treat is as OutOfMemory
• [Indexing] Reduce amount of IO needed to deal with temporary files
• [Indexing] Added ability to limit number of concurrently running indexes to reduce the memory and IO usage (Indexing.MaxNumberOfConcurrentlyRunningIndexes)
• [Indexing] Fixed invalid memory allocation in Lucene
• [Indexing] Added cluster-wide disable index
• [Indexing] Reduced logging lever of errors related to using Nuget DLLs in indexes
• [Indexing] Normalize new line characters when comparing additional sources
• [Fixed] Fixed the issue that we skip documents without checking if we need to check for CanContinueBatch which resulted in OutOfMemory error
• [Querying] Better exception on invalid JS Date
• [Querying] Ensured that query execution is stopped if a HTTP request is aborted
• [Querying] Streaming on index that does not exist should throw proper exception instead of NRE
• [Querying] Fixed NRE when using a search query with Boost on term containing only quotes
• [SNMP] Exposed number of disabled, encrypted and current node databases
• [SNMP] Exposed GC memory info
• [Server] Disabling Kestrel's AllowSynchronousIO and making the handling of requests fully async. Added Http.AllowSynchronousIO to allow user to control it explicitly.
• [SQL ETL] Added handling for conversion of GUID to UUID when loading records to PostgreSQL
• [Time Series] Fixed append of an entry with DateTime.MinValue
• [Time Series] Avoid throwing an exception when timeseries name isn't a string
• [Time Series] Added support for quering time series with quoted name
• [Tombstones] Added endpoints allowing to force cleanup of tombstones and check the cleanup state
• [Voron] Fixed unhandled case in page splitting of compressed pages. This applies to map-reduce indexes and could result in "Encountered empty page which isn't a root" error.
• [Voron] Ensured that we report an alert if an error happened during the flush of data and we recreate the backgroud task
• [Voron] Fixed unreleased allocations in scratch buffers which made that scratch files were never cleaned up
• [.NET Core] Updated to 5.0.5

Client

• [Changes API] Fixed detection for database creation/deletion while already suscribed
• [Session] Consider the index referenced collections on WaitForIndexesAfterSaveChanges
• [Conventions] Added WaitForIndexesAfterSaveChangesTimeout, WaitForReplicationAfterSaveChangesTimeout and WaitForNonStaleResultsTimeout conventions (default: 15 seconds)
• [Performance] Optimized TryGetServerVersion
• Fixed NRE in CachingJsonConverter
• Added Lazy version of ConditionalLoad
• IAsyncDocumentSession.StreamAsync method forcedly cast IAsyncDocumentQuery<T> to concreate AsyncDocumentQuery<T> class implementation

Embedded

• Added support for wildcards (x) in the ServerOptions.FrameworkVersion

Studio

• [Querying] Added id() field to the list of index terms

Server

• [Backups] fixed issue with deleting multiple blobs on Azure when retention policy is executed
• [Backups] fixed issue with executing retention policy on Azure when remote folder name is empty
• [Backups] fixed the ability to test credentials on server-wide backups
• [Certificates] set DisableCertificateDownloads to true when building a certificate chain
• [Cluster] fixed issue with excessive topology updates from the Client API
• [Cluster] do not move loading idle database to rehab
• [Compression] fixed issue with compressed flag not being added when number of overflow pages is exactly the same after compression
• [Debug] fixed issue with retrieving thread names in Threads View for some of the threads
• [Encryption] fixed 'unable to decrypt page' error
• [Encryption] fixed recovery process on encrypted storage which could cause data corruption
• [Graph API] do not return duplicate edges for Graph Visualizer
• [Indexes] memory optimizations in Lucene
• [Indexes] fixed race condition between removing or replacing an index and starting or stopping all of the indexes
• [Indexes] changing additional sources should affect index etag which is used to calculate query results caching etag
• [Indexes] expected exceptions should not generate an index error
• [Indexes] decreased the cost of retrieving the data from index stats endpoint
• [Indexes] fixed issue with disposing cached field items immediately - some of the readers could be already in use
• [Indexes] when document collection does not match the specified collection in LoadDocument then null should be returned
• [Indexes] index definition comparison should be newline-insensitive
• [Monitoring] exposed SWAP usage on Windows in SNMP
• [Monitoring] added managed memory, unmanaged memory, encryption buffers memory in use and encryption buffers memory in pool to SNMP
• [Patching] fixed issue with delete script not deleting all of the documents in some cases
• [Patching] fixed memory leak when patching a large collection on a compressed database
• [Querying] fixed possible NRE when projecting with streaming full documents
• [Replication] fixed NRE that could happen if there are 2 tombstones with identical ID but different collection
• [Replication] fixed NRE that could happen if we are deleting the database
• [Replication] properly clone replication items to avoid using memory not owned by the replication process which could result in AccessViolationException
• [Revisions] fixed issue with creating a revision when only Resolved flag was present in deleted document
• [Revisions] properly set flags during putting deleted revision from a cluster transaction
• [Subscriptions] fixed NRE in subscriptions when including counters that do not exist in the document
• [.NET] updated to 5.0.3

Client API

• [Operations] added 'GetRevisionsOperation'
• [Querying] wrap next binary expression after negation in sub clauses
• [Session] fixed issue with lazily load documents not using aggressive cache properly
• [Subscriptions] fixed ObjectDisposedException that could occur when disposing subscription worker multiple times
• [Subscriptions] fixed JavaScript translation for Id property and Load, allow to use member init
• [Subscriptions] do not redirect to node with failures

Studio

• [Backup] added 'Content-Type' header so browsers like Firefox can then recognize file name properly of exported database
• [Databases] do not show document and index counts on databases that do not belong to the current node (remote)
• [Indexes] fixed issue with Indexing performance view not loading correctly when cmpxchg are used in index

Server

• [Encryption] Fixing the recovery process on storage startup which could cause overwrite of valid data and result in "Unable to decrypt page" error
• [Encryption] Fixing potential segmentation fault error on Linux on attempt to encrypt empty page that effectively was not in use
• [Indexing] Fixing internals in Lucene to prevent from killing the server by an unhandled exception