Rhino Changelog

When deploying resource adaptors that use Rhino’s stats annotation classes, "unknown enum constant" javac warnings (although harmless) will no longer be output to the Rhino log. (#1758900)

MemDB migrating locks still present for removed keyspaces are now excluded from the extracted state stream used to resynchronise booting cluster nodes so they don’t cause a NullPointerException during state installation on the booting node. (#1685609)

Fixed issues with SNMP parameter set type registration behaviour when the SNMP agent cannot start successfully, and fixed SNMP agent alarms not being cleared when restarting the agent (#1757618)

Fixed a potential NullPointerException during convergence checks if service actual state is absent. (#1764911)

Fixed the missing node startup logging of an installed license if the license contained only wildcard versions and no function with a capacity of zero. Removed the output of a warning message from the getlicensedcapacity rhino-console command when a function version is actually covered by a license with a wildcard version for that function. (#1775284)

Fixed a copy and paste typo in the snmp.filename.max-length system property such that it was actually obtaining its value from the snmp.name.max-length system property instead. (#1859694)

Fixed a rare ConcurrentModificationException that could be thrown when a node looks up cluster node roles at the same time as the node enters or leaves the primary component. (#1864114)

Fixed a bug that caused the DirectStatsProxy’s IOLoop thread to hit 100% CPU usage. (#1747793)

Added a new watchdog to monitor for adverse GC events called GC Watchdog. This watchdog monitors GC activity and raises an alarm if a long GC pause is detected. Enabled only in pool mode. (#1713035)

MemDB migrating locks still present for removed keyspaces are now excluded from the extracted state stream used to resynchronise booting cluster nodes so they don’t cause a NullPointerException during state installation on the booting node. (#1685609)

Added commented-out properties to Rhino start scripts and security properties to simplify enabling X.509 Certificate revocation checks. (#1635614)

Added missing netty classes to cassandra support jar, required when using TLS connections to the database. (#1641997)

Fixed a threshold rule processor issue that caused rules using the same counter statistic in different modes (absolute vs. delta) to not evaluate correctly for one of the modes. Fixed a small memory leak when threshold rules were deleted. (#1620902)

Fixed issue with missing client utilities class that prevented creation of declarative exports. (#1535946)

Rhino now recognises deployable units signed using the RSA and ECDSA algorithms (in addition to the DSA algorithm it already supported). (#1570188)

The Rhino console now supports the escape sequences generated by Putty’s application numpad mode. (#1316710)

Disabled JVM assertions in a PostgreSQL JDBC driver class to work around a rarely-occurring bug in the driver code. (#1394784)

Update the org.postgresql:postgresql library to 42.7.2

This release of Rhino introduces support for Java 17 and has been tested with the Microsoft build of OpenJDK 17 (https://learn.microsoft.com/en-us/java/openjdk/download#openjdk-17). Existing support for Java 11 is unchanged.

Optimised several internal Rhino debugging code paths. (#1087376)

Removed a source of contention when enqueueing SAS events for transmission. (#1053165)

Removed repeated logging of a duplicate address warning in rhino-stats when multiple Rhino nodes report the same JMX IP address, which can occur when Rhino nodes are running in containers. The warning is now reported only once. (#830359)

Fixed the rhino-stats client so that CGIN stats can also be collected when the wrong version is specified in the file defining the stats to be collected. (#1067916)

Fixed a bug where the rhino-stats client could not collect stats if the client was on RVT version 4.1 while the server was on 4.0. (#1068980)

Fixed a bug where cancelling a remote timer that was set on a recently restarted pool node could fail. (#1127179)

Update the netty libraries to 4.1.103.Final

The stats collection feature can now continue to collect stats even if the wrong versions are specified in the file defining the stats to be collected. This means that MAG nodes no longer need to be restarted after upgrades of RVT nodes. (#751400)

Update the SLEE Annotations dependency to 3.1.4

Rhino SDK and Production now use the Java G1 garbage collector by default. (#443219)

Fixed a rhino-stats bug where quiet mode (-q) used in conjunction with rolling stats output mode (-o) would prevent any stats being written to the output file. Also improved the command line help description for rolling file output mode. (#826721)

Fixed a bug where a lock manager statistics parameter set may be unavailable for reading statistics after undeploying and then redeploying the same SLEE service or profile specification. (#923567)

Fixed a rare race condition where undeploying and then redeploying a SLEE service or profile specification could result in the associated lock manager incorrectly removing locks. (#915873)

Added tolerance to the rhino-stats collection so that available stats will get collected even if any of the selected stats were unavailable. (#751400)

Added a 5s timeout to RMI client connection health checks as a workaround for rare connection hangs in JDK library code. (#835614)

A maximum file size in terms of code points is now enforced for declarative YAML documents. The default is 10,485,760 which may be increased or decreased using the system property rhino.config.yaml_code_point_limit. (#920468)

When running in pool mode, Rhino nodes will automatically restart if an OutOfMemoryError is encountered. (#751195)

Updated org.yaml:snakeyaml to 2.0.

Updated org.postgresql:postgresql to 42.6.0.

Updated com.google.protobuf:protobuf-java to 3.19.6.

Updated io.grpc libraries to 1.58.0.

Updated org.xerial.snappy:snappy-java to 1.1.10.5.

When running in pool mode, management functionality that is not relevant for this mode has been disabled.

Removed the PoolMembershipFacility API in favour of an enhanced ClusterStateChangeFacility instead. Adding pool mode support directly to the ClusterStateChangeFacility allows resource adaptors that depend on the types of notification generated by this facility to continue working with approximately the same behaviour, without any code changes, when Rhino is running in pool mode. (#370019, #368243)

Support for clients obtaining Rhino statistics using the direct connection mode is now optional and can be disabled if desired (previously it was always enabled). The network interface that direct connection sockets are bound to in Rhino can now also be configured in rhino-config.xml New questions have been added to the production installer to determine the initial deployment configuration, with the install defaults being the same as set in previous Rhino versions. (#665937)

Support for allowing profile snapshots by management clients is now optional and can be disabled if desired (previously it was always enabled). Configuration for the snapshot server has moved from a runtime system property to rhino-config.xml. New configuration options for the snapshot server have been added, including allowing the network interface that server sockets are bound to to be specified. New questions have been added to the production installer to determine the initial deployment configuration, with the install defaults being the same as set in previous Rhino versions. (#665934)

Fixed a rare race condition that allowed resource adaptor entities to be created by convergence tasks (as a result of declarative config import) while the actual SLEE state was starting, which later led to an AssertionError when the actual state of the SLEE changed again. (#822565)

Fixed issue with rhino-export -R producing an invalid export when writing to a destination path containing multiple path segements. (#474399)

Fixed bug where rhino-console would ignore the password command line parameter (-w) if it appeared before the username (-u). (#731287)

Fixed a bug in rhino-console where the port command line argument (-p) would be ignored if the rhino.remote.serverlist property is defined in client.properties. (#366785)

Fixed a memory leak when direct stats connections between cluster nodes, used for reporting local stats to the session-controlling node, are closed after a stats session ends. (#500827)

Fixed precedence of the rhino-stats command line connection options for host (-h) and port (-p) so they override any serverlist or adhoc addresses specified in the client.properties file. (#270120)

The rhino.remote.serverlist property is no longer used by rhino-stats when creating connections to pool nodes. (#319909)

When running in pool mode, the rhino-stats client will no longer make multiple connections to the same JMX endpoint address when multiple nodes report using the same address. Instead, duplicate addresses will be ignored. (#220533)

Fixed a bug in rhino-stats where identifiers set with the -I option were not quoted in CSV output despite containing commas. (#441687)

The rhino-stats client will no longer display the previous value for a stat if there is no new value for the stat in the latest update from Rhino. (#384080)

Fixed a NullPointerException that occurred in the rhino-stats GUI when double-clicking on namespace grouping nodes in the left-hand side parameter set tree. (#646650)

Fixed a NullPointerException that occurred in the rhino-stats GUI if a parameter set with no parameter set type is selected in the "Select Parameter Set" dialog box (when adding new statistics to an existing graph). Also fixed the enabled/disabled behaviour of the "OK" button in this dialog box so that it is only enabled when at least one statistic is selected. (#646742)

Fixed some session ownership store alarm names to refer to session ownership rather than the key/value store. Renamed the "table-missing" alarm names for the key/value store, session ownership store, and pool maintenance provider, to "missing-table" for consistency with the "missing-keyspace" alarms. (#376162)

Missing keyspace or table alarms raised as a result of runtime query failures will now be correctly cleared when the problem is resolved. The alarm instance identifier of missing table alarms raised as a result of runtime query failures has also been corrected (they were not being qualified with the associated keyspace name). (#511248)

Added a schema-change listener to the Cassandra database sessions used for persisting resources (key/value store, session ownership store, and pool maintenance provider). This means that if a database keyspace or table used by one of these resources is manually dropped from Cassandra then the resource will notice immediately and take appropriate action - either recreate the missing keyspace and/or tables if data definition updates are allowed, or raise an alarm if not. As a side effect, this fixes related issues such as the pool metadata table not being repopulated if the metadata table is dropped and recreated. (#530772:)

Fixed the cleanup of replicated storage keyspaces when a resource adaptor entity using them is removed when the keyspace is replicated using the key/value store. This bug also caused a discrepancy between nodes about which Cassandra tables should and should not exist. (#721572)

The alarms raised by persisting resources (key/value store, session ownership store, and pool maintenance provider) are now included in the Rhino alarm catalog. (#378603)

Exceptions from the com.opencloud.rhino.resource package thrown by subtypes of PersistingResourceMBean are now rendered correctly by remote clients. (#386568)

Additional exceptions that can be thrown by gRPC services (remote timers, pool mode message facility) are now caught and handled appropriately. (#370127)

Fixed a NullPointerException that could occur during convergence is a resource adaptor entity was removed while the SLEE was starting.

Added missing javadoc in the install packages for key/value store, session ownership store, and pool maintenance subsystem. (#426216)

Fixed a bug where errors in the set up of the rhino-stats collection caused a memory leak. (#751399)

The log4j logger configurations representing tracer levels for services and profile tables will now be updated during desired state convergence checks if it’s determined that the logger configurations do not match previously configured tracer level desired state. (#644804)

The logging polled memory appender now supports CIDR notation for its bind address, allowing a network range to be specified. The appender will bind to the network interface it finds within that range. An error will occur if exactly one network interface cannot be found within the range. (#665936)

Added a new resource adaptor extension configuration property to RhinoExtensionProperties that returns the local Rhino node ID as a java.lang.String. (#260512)

Static OIDs have been assigned to the stats parameter set types of the key/value store, session ownership store and pool maintenance provider. (#421461)

The rhino-stats client now logs warning and error messages to STDERR instead of mixing them with output in STDOUT. (#384075)

The upper limit of the rhino-stats pool period config has been removed. (#375849)

Added new parameter sets for Rhino interconnect gRPC client and server statistics. (#235852)

Added a new counter statistic to Cassandra resources that records any internal exceptions raised while handling result sets from a query operation. (#379164)

Replaced the Cassandra persisting resource driver timeout and driver error alarms with specific logging and statistics. This prevents driver related alarms from remaining uncleared after the issue is resolved. (#376653)

The pool mode MessageFacility implementation now imposes a timeout for responses to sent messages. (#385131)

IPv6 address are now supported by the Rhino client connection library.

The Cassandra Datastax driver reference configuration is not included in the Rhino install documentation subdirectory. (#422004)

Added more restrictive permissions for temporary files. (#412308)

Remote timer internode communication now uses the default interconnect request timeout configured in rhino-config.xml (default 5s) rather than a separately configured system property (default was 2s). (#396582)

Improved error messages reported when attempting to invoke scattercast management operations when scattercast is not configured. (#262419)

Updated the LDAP login module to use TLS by default. (#412311)

Fixed known issues in Cassandra resources with missing keyspace and table alarms not being cleared when a corresponding keyspace or table is no longer required due to deployment changes. All alarms related to a particular Cassandra persistence instance are also now cleared if that persistence instance is removed from a persistence resource. The clearing of Cassandra driver timeout alarms is now managed on a per-session basis rather than globally, i.e. a timeout alarm raised for a particular session will now be cleared after the configured period if there are no more timeouts reported by that session, irrespective of timeouts that may occur on other database sessions. (#370627)

Fixed alarms for failures in Cassandra resources where missing table alarms were not raised in certain failure cases. (#220529)

Fixed hardcoded appender ref modification problems. Logging will now continue to the audit logs after an appender ref is added. The additivity of the logger will now correctly have an additivity of false. (#377792)

Fixed the runtime exception thrown by the MessageFacility when trying to send broadcast messages when using Rhino Pools mode. (#370122)

Fixed the parsing of a configured SNMP notification target to allow hyphens in the middle of the hostname. (#423328)

Updated com.fasterxml.jackson artifacts to 2.14.1.

Updated org.apache.common:commons-text to 1.10.0.

Improved reliability in various failure cases. Restrictions of quorum and a distributed consistent view between cluster members are removed. Rhino nodes are independent so if more than half of the nodes simultaneously fail the remainder continue to function.

No distributed locking amongst the Rhino nodes.

A more permissive node liveness model. Rhino nodes will remain available in more cases of solution component failure.

Increased scalability. The maximum size of a Rhino Pool is much higher than a Savanna based Rhino cluster.

The Savanna transport is used for the message facility

The new node interconnect is used for remote timers

Added exportpersistingresourcedatadefinitions rhino-console command and associated MBean method. This command is used to export Cassandra resource database schemas for importing into a Cassandra database. (#185430)

Cassandra persisting resource keyspace names may now be configured at install time. (#185433)

Added dumppersistingresourcetable rhino-console command and associated MBean method. This command allows dumping persisting resource (i.e. database) contents for diagnostic purposes. (#222039)

Added listpersistingresources rhino-console command and associated MBean method. This command lists all configured persisting resources. (#222039)

Improvements to the rhino-stats client to allow collection of stats from Rhino Pools and from nodes in different clusters in an 'adhoc' mode. (#256356) When run in adhoc mode the nodes' addresses are added as a comma separated list of host:port (JMX MBean server addresses) either:

Modified Rhino’s internal handling of MemDB keyspaces to allow for the keyspace names to be stabilized. The keyspace names can be reused and generated consistently. (#338386)

Added flags to disable execution of DDU queries (data definition updates). This enables deployments where the Cassandra installation is managed separately from the Rhino cluster with restrictive access controls or when Rhino is deployed in pool mode. This behavior is configurable at install time, the default is to automatically update schemas unless using pool mode where automatic updates are prohibited. (#185432)

The remote timer CompositeData and TabularData MBean objects now return the remaining repetitions of a timer with the repetitions field, instead of the expiry time with the expiry-time field. (#334324)

Fixed node restart failure when deployed library components are referred to by other SLEE components that have been verified but not deployed. (#358369)

Removed client fallback to insecure JMX connections if SSL negotiation fails. The Rhino management interface does not support insecure connections. (#195604)

Fixed a sporadic management lock timeout when shutting down the SNMP agent during reconfiguration. (#360655)

Fixed issue where sample stats distributions were calculated between the bin distribution min and max, which could lead to odd stats (most notably negative percentile stats) if min < observed min or max > observed max. (RHI-3505)

Fixed MemDB retained size accounting when the key for a deleted but not yet GC’d entry is reused. (#337944)

Added gRPC 1.47.0 and associated dependencies.

Java 11.0.15 is now the minimum supported version. This version disables the insecure TLS-1.0 and 1.1 protocol versions in the SSL socket implementation.