Minor Releases Enterprise Edition
This page provides a dedicated overview of patch releases for the Axon Server (Enterprise Edition) releases
Fix: Improve loading aggregate events performance for clients connected to a follower
Fix: Failure in the initialization of a node in leader role now prevents the node from moving to leader role
Fix: Message from another node with a higher term must force node into follower state if it is in leader state
Fix: Improved error handling and feedback when uploading invalid plugins
Fix: Increase default settings for spring.servlet.multipart.max-request-size and spring.servlet.multipart.max-file-size to 25MB
Fix: Close event store segment file when reading is complete
Fix: In case of timeout during query execution, AS sends a timeout error to the client before canceling the query
Fix: Queries and commands not cancelled after a timeout
Added type (Event/Snapshot) as a tag for metric indexes on JumpSkipIndex
Fix for contexts not starting when some contexts are using BLOOM_FILTER_INDEX and others are using JUMP_SKIP_INDEX.
Fix: Clients should not connect to a node that is unable to contact other nodes in the cluster
If a client requests a connection from an isolated Axon Server node, Axon Server rejects the connection, and the client
requests a connection from another node. If a client has a connection to an Axon Server node, and the node loses the
connection to the other Axon Server nodes in the cluster, Axon Server will disconnect the client. The client reconnects to another
node (also in 4.4.16).
Fix: Regression in loading aggregate events performance
Fix: Handle queries with same request type but different response type
Fix: client sometimes reconnects when other Axon Server node is restarted
Fix: clients are redirected to Axon Server node that is not ready yet
Fix: invalid version number in login and error pages
New metrics added:
file.bloom.open: counts the number of bloom filter segments opened since start
file.bloom.close: counts the number of bloom filter segments closed since start
file.segment.open: counts the number of event store segments opened since start
local.aggregate.segments: monitors the number of segments that were accessed for reading aggregate event requests
Notes:
Default value for configuration property axoniq.axonserver.event.events-per-segment-prefetch is decreased from 50 to 10.
Improved performance for reading aggregates
Axon Server is now reading events for an aggregate from multiple event store segments in parallel. The order in which Axon Server returns the events remains unchanged.
Reduced memory usage for in-memory indexes
Axon Server maintains index entries for the latest event store segment in-memory. The structure of this data has been changed to reduce the heap used by this index.
Reduced number of files kept open as memory-mapped files
Axon Server now only keeps a (configurable) number of event files open as memory-mapped files. Older files will be opened and closed when needed. The number of event store segments that are opened as memory mapped files can be configured through the property: axoniq.axonserver.event.memory-mapped-segments
Initialize event stores asynchronously on startup
When Axon Server needs to recreate indexes for a context on startup, it can take some time to complete. This change makes the initialization of the context asynchronous, so that if one context takes a long time to initialize, other replication groups are already available. This also makes the HTTP endpoint available sooner, so liveliness checks succeed faster.
Improvements in shutdown process
Fix: Load balancing operations for processors should ignore stopped instances
Fix: Stop reading events when query deadline expires
Fix: Disparities in Context Leaders
Dependency update: updated xstream version used to 1.4.17
Configurable strategy for aggregate events stream sequence validation (through property read-sequence-validation-strategy)
Fix UI check for updates
Fix for initialization error for the transaction log
Fix for node not able to replicate after install snapshot procedure
Fix for concurrency issue in creating schedulers for auto-loadbalancing
Fix: Failure in the initialization of a node in leader role now prevents the node from moving to leader role
Fix: Message from another node with a higher term must force node into follower state if it is in leader state
Fix: Clients should not connect to a node that is unable to contact other nodes in the cluster
If a client requests a connection from an isolated Axon Server node, Axon Server rejects the connection, and the client
requests a connection from another node. If a client has a connection to an Axon Server node, and the node loses the
connection to the other Axon Server nodes in the cluster, Axon Server will disconnect the client. The client reconnects to another
node.
Fix: Load balancing operations for processors should ignore stopped instances
Fix: Stop reading events when query deadline expires
Dependency update: updated xstream version used to 1.4.17
Configurable strategy for aggregate events stream sequence validation (through property read-sequence-validation-strategy)
Fix for initialization error for the transaction log
Fix for node not able to replicate after install snapshot procedure
Fix for concurrency issue in creating schedulers for auto load balancing
Fix for race condition during JumpSkipIndex update
Fix for memory leak during peer node registration in Raft leader
Fix for subscription queries in case of missing query handler
Fix for concurrency issue in listing aggregates events during appending events for the same aggregate
Load balancing strategy for queries is now configurable. Metrics based by default, to configure round-robin set property
axoniq.axonserver.query-handler-selector=round-robin
Improvement on metrics based load balancing of queries, now giving higher probabilities based on the response time
for the handler as recorded on the current axon server node
Fix for scheduler not rescheduling all events after a leader change
Improvement for subscription query: initial result are now provided by a single instance per component
Fix for log compaction never performed if the backup is taken more often than each hour
Fix in release notes for version 4.4.7
Fix for processor information showing information on disconnected applications
Fix for issue with null expressions in ad-hoc queries
Updated GRPC version to 1.34.0
Added option to limit the number of commands/queries in progress per context
Fix for access control issue when updating an application
Fix for global index files remaining after delete context
Fix for commit index not updated correctly in specific cases
Fix for warning messages on opening index that should have been debug messages
Fix for invalid previous term/index in replication heartbeat messages causing some overhead in communication
Fix for missing portnumber in cluster template export
Fix for commands being sent to wrong context if the only handler is in another context
New property "axoniq.axonserver.enterprise.default-index-type" to specify the index type for new contexts. Default
value is JUMP_SKIP_INDEX, use BLOOM_FILTER_INDEX for the pre-4.4 index type
Improved QueryService logging
Added preserve event store option to delete context CLI command
Fixed stream completed by the server in case of inactivity
Hide upload license panel in SE
Reduced number of open index files
Fix for GetTokenAt operation
Improved feedback on license upload errors
Fix timing issue in leader change potentially causing duplicate events
New REST endpoint to download a diagnostics zip file
Fix for connections not correctly registered
Changed initialization sequence for event store to initialize completed segments first
Changed order of files in the backup endpoint for contexts with a jump skip index to list the global index files
before the segment index files
Fix timing issue when a follower sends new events to a new leader before it is fully initialized
Improved logging and error handling for log compaction task
Fix for initializing jump skip index when it has more than 1 segment
Improved error handling for problems creating a context and storing events/snapshots
Offload expensive data-writing operations to separate thread pool
Fix for reading aggregates with older snapshots
Fix race condition in queries and commands handlers unsubscription during reconnection
Fix pre-vote election with active backup nodes
Axon Server SE improvements from 4.4 to 4.4.1
Fixed issue causing added latency while Tracking live Events from a follower node
Fix the event processor status refresh process
Fix for downloading and starting with cluster templates
Renamed field replicationsGroups to replicationGroup in cluster template
Fix storing entries when a context is created with pre-existing event store
Fix query from dashboard when query is executed from admin node and the target context is not defined on this admin node
Fix for timing issue in sharing metrics between Axon Server nodes causing exception during delete context
Fix for migration of contexts created before version 4.3
Fixed concurrency issue in subscribing/unsubscribing commands
Do not override log entries in RAFT log that already have been committed
Fix for scheduling of tasks in clustering module
Do not override log entries in RAFT log that already have been committed
Fixed logging in IndexManager
Reduced risk for contention when opening an index file
Offload expensive data-fetching operations to separate thread pool
Option to configure the way that index files are opened (memory mapped or file channel based)
Limit the amount of commands/queries held in Axon Server waiting for the handlers to be ready to handle them, to avoid
out of memory errors on Axon Server
Fix for high number of cluster-request threads being created
Fix for timing issue in delete context. This could leave the context existing on one of the member nodes
Fix RAFT bug: configuration changes are not allowed before an entry has been committed in the current term.
New configuration properties added for Axon Server:
axoniq.axonserver.data-fetcher-threads=24 (number of threads that are allocated for doing longer running operations on the event store)
axoniq.axonserver.command-queue-capacity-per-client=10000 (number of command requests for a specific command handling client that Axon Server will cache waiting for permits)
axoniq.axonserver.query-queue-capacity-per-client=10000 (number of query requests for a specific query handling client that Axon Server will cache waiting for permits)
axoniq.axonserver.replication.use-mmap-index=null (by default, AxonServer will determine whether to use memory mapped indexes for replication logs based on operating system and java version, in rare cases it may be useful to override the default)
axoniq.axonserver.replication.force-clean-mmap-index=null (option to forcefully close unused memory mapped files instead of leaving the garbage collector do this, by default, AxonServer will determine this based on operating system and java version, in rare cases it may be useful to override the default)
axoniq.axonserver.event.use-mmap-index=null (by default, AxonServer will determine whether to use memory mapped indexes for event files in the event store based on operating system and java version, in rare cases it may be useful to override the default)
axoniq.axonserver.event.force-clean-mmap-index=null (option to forcefully close unused memory mapped files instead of leaving the garbage collector do this, by default, AxonServer will determine this based on operating system and java version, in rare cases it may be useful to override the default)
axoniq.axonserver.snapshot.use-mmap-index=null (by default, AxonServer will determine whether to use memory mapped indexes for snapshot files in the event store based on operating system and java version, in rare cases it may be useful to override the default)
axoniq.axonserver.snapshot.force-clean-mmap-index=null (option to forcefully close unused memory mapped files instead of leaving the garbage collector do this, by default, AxonServer will determine this based on operating system and java version, in rare cases it may be useful to override the default)
Configuration properties default values changed:
axoniq.axonserver.cluster-executor-thread-count=4 (reduced from 8, configures the number of threads used to handle requests from other axon server nodes, reduced as effective processing of the requests if offloaded to other threads)
axoniq.axonserver.executor-thread-count=4 (reduced from 8, configures the number of threads used to handle requests from clients, reduced as effective processing of the requests if offloaded to other threads)
Fix for race condition during rollback of transaction log
Fix for tracking event processor updates to websocket causing high CPU load in specific situation
Reduced warnings in log file on clients disconnecting
Fix for concurrency issue in sending heartbeat while client connects/disconnects
Updated usage output in CLI
Updated gRPC/Netty versions
Prevent errors in log (sending ad-hoc result to client that has gone, sending heartbeat to client that has gone)
Fix in CLI, application list failed
Removed retry option from H2 controldb URL
Fixed Axon Server keeping incorrect application references on restart of Axon Server node where application was connected
Fixed race condition when two Axon Server nodes are trying to become member of the same context at the same time
Fix for some REST reads were able with invalid token
Clean up of server shutdown process
Fixed error when adding a non-existing node to a context
Rollback in log entry could cause cluster member to transition to fatal state when there were writes pending
Fix for start-up issue when management.server.port is set
Added instruction acknowledgements
Client applications heartbeat support
Cleaned-up logging
Fix for specific error while reading aggregate
Optional heartbeat between Axon Server and Axon Framework clients
Fix for issue on cancelling queries on lost connection
Introduced acknowlegments on instructions between Axon Server and clients
Fix for continuously checking tracking event processor status
Fix for issue in access control: combining roles for all contexts and single context
Fix for issue in access control: applications with access to multiple contexts are not always correctly validated
Fix for UI issue: after deleting an application the buttons no longer work
On auto-load-balance of a event processor group only request the current status from applications with this processor group
Fix for situation where election takes too long to complete causes 2 leaders
Wait for leader when event store requests arrive at Axon Server during leader election. New configuration property
axoniq.axonserver.replication.wait-for-leader-timeout to change this wait time. Default is 2*max election timeout.
Fix for unexpected aggregate sequence exception.
Fix for wrong version number in dashboard.
Fix for timing issue when new events arrived at new leader before leader was fully initialized
Fix for configuration not always applied on all nodes
Fix for OutOfDirectMemory issue when installing snapshot
Fix for unexpected leader change on adding 3rd node to context
Fix for replication entry log compaction
Fix for mismatch in connected applications after rebalance
Fixes in access control check on REST endpoints for events/snapshot endpoints and renew application token
Improved error handling in replication
Fixed occasional timing issue in create context
Improved recovery options after full data loss on node
Fix for authorization path mapping
Fix for subscription query memory leak
Improvements in error reporting in case of disconnected applications
Improvements in detection of insufficient disk space
Fix on RAFT Leader detection
Avoided multiple concurrent creation of the same context
Fix for appendEvent with no events in stream
Stop log cleaning task when context is deleted or node is removed from context
Fix for NullPointerException on lost connection
Improved recovery of cluster after disk failure
Fix for updating configuration during install snapshot
CLI commands now can be performed locally without token.
Improvements in replication to nodes that have been down for a long time
Tracking event processor auto-loadbalancing fixes
Status displayed for tracking event processors fixed when segments are running in different applications
Tracking event processors are updated in separate thread
Logging does not show application data anymore
Fixed Axon Dashboard list of cluster nodes missing hostname and ports
Changed some gRPC error codes returned to avoid clients to disconnect
CLI commands init-cluster/register-node/add-node-to-context are now idempotent
Register node returns faster (no longer waits for synchronization to new node)
Reduced risk of re-election if a node restarts
Fixed occasional NullPointerException when client connects to a newly registered Axon Server node
Fixed incorrect leader returned in context API and multiple leaders warning in log
Default controldb connection settings changed
gRPC version update
Register node no longer needs to be sent to leader of _admin context
Merge tracking event processor not always available when it should
Logging changes
Fix for queries timeout
Fix for replication with large messages
Added axonserver-cli.jar to release package (axoniq-cli.jar is deprecated)
