OCSS7 Changelog

The SGC will no longer use stale routing data when a transaction ID is reused following termination of the previous incarnation of that transaction ID due to receipt of UDTS or XUDTS. (SSSVN-1247)

Improved the robustness of the SGC’s handling of TCAP stack connect and disconnects. (SSSVN-1068)

The PeerManagerObjectManager is now more robust if unexpected Hazelcast data loss occurs. (SSSVN-1063, SSSVN-1117)

DPCInfoHelper should no longer sometimes throw NullPointerException following route deletion. (SSSVN-1113)

The CTS-SGC-MIB should now correctly validate, with the exception of two object type names: sgcSccpErrorStatsEntryNoTranslationForAddressOfSuchNatureReceived and sgcSccpErrorStatsEntryNoTranslationForThisSpecificAddressReceived. (SSSVN-1124)

Corrected defects that could result in stuck dialogs and increased Rhino CPU after mass dialog state loss. (SSSVN-1038)

The SGC will now accept malformed M3UA messages that are missing their final parameter padding octets. (SSSVN-787)

Fixed SNMP’s misreporting of the following stats, which could appear with a higher value than the actual count when read. This affected the SNMP interface only, CLI display of these stats was always correct. The fixed stats are, for SCCP: segmentationFailureCount, unitdataIndCount, reassembledUnitdataCount, reassemblyFailureCount; and for Health: forceAllocatedReqTasks. (SSSVN-789)

Fixed an issue that could cause the SGC to restart if more than 1 million transactions were being concurrently handled by a single TCAP stack connection. (SSSVN-868)

CGIN, SIS and the IN Scenario Pack now support ANSI SCCP. Note that this requires the usage of an SGC from OCSS7 2.1 or higher.

Fixed defect where TCAP stack could hang if stopped while dialogs were still active. (SSSVN-779)

Fixed a deadlock that could occur during TCAP stack deactivation. (SSSVN-780)

Fixed a defect that prevented sending TC-ABORT in response in incoming TC-BEGINs while the TCAP stack was deactivating. (SSSVN-781)

Fixed a double-free in the TCAP stack which caused transactions to be reallocated whilst still alive, resulting in permanent entries in CGIN’s internal dialog maps. (SSSVN-1035)

Corrected an issue that could result in an AccessControlException if finest logging was enabled in the OCSS7 TCAP stack. (SSSVN-889)

Upgraded Hazelcast to version 3.7. (SSSVN-350)

The SGC is now able to detect when distributed Hazelcast data has been irrevocably lost and will restart the node in order to load configuration from file. (SSSVN-357)

Added more statistics: TcapStats, SccpStats, TcapErrorStats, SccpErrorStats. These are accessible via the command line client, JMX and SNMP. Some statistics previously in LocalSsnInfo are now recorded in SccpStats instead. (SSSVN-668)

The SGC will now detect and apply log4j.xml changes in real-time. (SSSVN-163)

Added a status field to display-info-tcapconninfo to display connection state. (SSSVN-426)

An alarm will be raised if the SGC detects that its default map backup-count configuration is considered too small for the size of cluster. (SSSVN-683)

SGC version information is now logged in the startup log and also on rollover in the main ss7.log. (SSSVN-737)

The SGC’s SNMP system OID is now 1.3.6.1.4.1.19808.10.2. (SSSVN-445)

Heartbeat is now enabled by default on all TCAP stack to SGC connections. (SSSVN-486)

The SGC now raises an alarm at MINOR if one or more paths in an association go down. The existing alarm for the whole association going down is now raised at MAJOR instead of MINOR. (SSSVN-715)

Memory requirements to support prefix migration have been drastically reduced in both the SGC and the TCAP stack. (SSSVN-732, SSSVN-733)

Local SSN prohibited alarm is now raised at MAJOR level. (SSSVN-738)

Demoted "Route routeId=x is DOWN - sending message back to SCCP" message from WARN to DEBUG. (SSSVN-143)

Fixed statistics problem where messages could be incorrectly recorded against SSN 0. (SSSVN-144)

Fixed issue that could generate a spurious "peer FSM exists in state: CLOSED will be replaced" message. (SSSVN-166)

Fixed NullPointerExceptions that could be generated if a SNMP node was configured with an invalid host name. (SSSVN-173)

Fix NullPointerException seen after disabling a local endpoint. (SSSVN-219)

Tab completion for 'display-info-ogtinfo' no longer suggests a non- existent 'ssn' column. (SSSVN-221)

Improved the error message displayed when the CLI batch mode cannot parse the provided batch file. (SSSVN-224)

TCAP stack connection loss no longer logs stack traces at WARN. (SSSVN-225)

Fixed an IllegalArgumentException seen when SNMP alarm IDs wrapped around. (SSSVN-356)

Fixed an AssertionError in Transport$ClientInitializer.initChannel() that occurred when an SGC was unable to connect to other cluster members via the comm switch interface. (SSSVN-362)

Fixed defect in decoder for SCCP importance field that could result in the wrong value being decoded. (SSSVN-385)

The CLI will no longer generate a "java.util.ArrayList cannot be cast to javax.management.Attribute" message when an unknown argument is passed to a display-info-xxx command. (SSSVN-394)

The SGC’s default minimum heap size is now equal to its default maximum heap size. The default minimum perm gen size is now the same as the default max perm gen size. (SSSVN-404)

STANDALONE hazelcast group name generation is now less prone to collisions when starting multiple standalone SGCs simultaneously. (SSSVN-425)

The SGC node manager now binds its listen socket with SO_REUSEADDR. (SSSVN-434)

The SGC no longer generates a NullPointerException if asked to send to a null destination SCCP address. (SSSVN-448)

The SGC will now attempt to restart if it exits with code 78. There is now a 30s wait between restart attempts. (SSSVN-514)

The CLI batch mode should now be significantly faster. (SSSVN-628)

Fixed an AssertionError in TcapRouter.globalRouteSelector that could occur if the SCCP reassembly timer raced with an incoming XUDT message. (SSSVN-660)

The Comm Switch will no longer be left in a zombie state if it fails to connect to other cluster members on startup. (SSSVN-681)

Fixed a NullPointerException in nodeLeft that could be generated when a cluster member left the cluster while another node was starting up. (SSSVN-682)

Fixed a defect where the TCAP stack would stop allocating new dialog ID following prefix wraparound. (SSSVN-691)

Fixed a prefix leak in the SGC that could result in the SGCs returning "out of prefixes" when a TCAP stack connected. (SSSVN-701)

Corrected an issue where a TCAP to SGC connection could end up not being used for any messages at all. (SSSVN-703)

Fixed a hard to trigger meshed connection manager deadlock. (SSSVN-704)

Under certain prefix migration conditions data structures associated with migrations would not be freed, resulting in an OutOfMemoryError. This has been corrected. (SSSVN-705)

Fixed TCAP stack leak that could occur if invoke timeouts expired while there were no connections available to any SGC. (SSSVN-712)

Corrected an issue where the TCAP stack could get its internal state confused during prefix migration. (SSSVN-713)

Corrected tab completion for display-info-tcapconninfo migratedPrefixes and tcapStackID columns. (SSSVN-734)

Installed a sensible default column display order for display-info- tcapconninfo. (SSSVN-735)

Added ability to connect to multiple SGCs simultaneously in a mesh style using new ocss7.sgcs TCAP stack property. (SSSVN-388)

SGCs now support failover of dialogs between connections to the same TCAP stack when using the new ocss7.sgcs TCAP stack property. (SSSVN-258)

sgc-cli.sh script now attempts to auto-detect SGC JMX host and port if those parameters are not set on the command line (-h -p) (SSSVN-292)

DAUD should no longer be slow to happen when many ASes are configured. (SSSVN-8)

Fixed an issue where IPv6 addreses in host:port format could not be parsed in the ocss7.urlList TCAP stack configuration property. (SSSVN-133)

"We should never have invoke timeout in UNUSED or ALLOCATED state" TCAP stack message has been downgraded from SEVERE to DEBUG. (SSSVN-182)

Segmentation configuration parameters MSS and MUSS will no longer permit unacceptable combinations. (SSSVN-209)

The CLI now reports a connection error rather than an unknown error when unable to communicate with the SGC. (SSSVN-228)

Fixed a ConcurrentModificationException that could be thrown in TcapRouter. This also fixes an issue where the TcapRouter could prevent graceful shutdown from completing. (SSSVN-315)

Fixed issue where "releasing begin prefix which has apparently been not assigned - no prefix at all" could be erroneously logged. (SSSVN-316)

Fixed very small thread local leak seen when rebuilding outbound translation data. (SSSVN-338)

Fixed an issue where SctpManagementMessages that were queued for transmission when a socket was closed were not released properly. (SSSVN-353)

sgcd script now creates /var/lock/subsys/sgcd entry. (SSSVN-365)

If the SGC fails to start up due to an Exception the stack trace from that Exception will now be logged at WARN (used to be logged at DEBUG). (SSSVN-376)

Made some usability fixes to the display-event-history command. (SSSVN-383)

SCCP defers registering to handle distributed tasks until after ReqCtx pools have been initialized. This prevents a NullPointerException at SccpManagement.sendSCMGMessage. (SSSVN-457)

Fix NullPointerException if an unknown alarm is raised via Hazelcast. (SSSVN-464)

Unexpected exceptions thrown while processing SGCTopics will now cause the SGC to attempt a graceful shutdown, similar to uncaught exceptions thrown elsewhere. (SSSVN-513)

Fixed NullPointerException in NodeObjectManager.nodeLeft when a node without configuration (create-node) leaves the cluster. (SSSVN-538)

Updated documentation to include Hazelcast configuration recommendations for clusters with more than 2 members. (SSSVN-550)

Applied a workaround for Hazelcast issue where a lock owned by an SGC could be unlocked underneath it during exit of another cluster member, generating "java.lang.IllegalMonitorStateException: Current thread is not owner of the lock!" messages. (SSSVN-604)

Fixed NullPointerException during config save. (SSSVN-626)

Fixed case where create-XXX could return before XXX had completed creation, resulting in subsequent enable-XXX or remove-XXX commands failing to notice that XXX was created. (SSSVN-626)

Decoder now correctly decodes Reject components that have no argument when received with component length in indefinite length form. (SSSVN-638)

Added -XX:+PrintGCDateStamps JVM flag to SGC startup script. (SSSVN-407)

Added documentation suggesting appropriate ulimit settings for user processes. (SSSVN-575)

Changed some default ports so that they’re no longer in the emphemeral range: SGC’s JMX port, Hazelcast multicast UDP discovery port. (SSSVN-618)

Fixed issue that could result in the SGC’s routing taking longer than expected to be available. (SSSVN-608)

Prevent NullPointerException during alarm unregistration while an SGC cluster split/merge is in progress from restarting the SGC. (SSSVN-553)

Failure of the comm switch to bind its listen port will now raise an alarm at critical level and attempt to rebind that port at regular intervals until successful. (SSSVN-531)

Corrected issue where global title translation tables were only updated on a single node after a local SSN status change. (SSSVN-536)

Fixed a ConcurrentModificationException that could occur when multiple threads attempted to modify the same configuration object simultaneously. (SSSVN-511)

Fix NullPointerException that could occur if global title rules were created and deleted very quickly. (SSSVN-420)

Corrected issue where display-info-remotessninfo was not updated cluster-wide on a local SSN state change. (SSSVN-517)

Fixed an issue which caused the SGC to leave gracefully disconnected TCAP stack connections in CLOSE_WAIT state indefinitely. (SSSVN-488)

Unexpected exceptions thrown while processing SGCTopics will now cause the SGC to attempt a graceful shutdown, similar to uncaught exceptions thrown elsewhere. (SSSVN-513)

Fixed an error where certain DPCs in SST/SSA/SSP messages would be incorrectly decoded. (SSSVN-314)

Fixed a race condition that could result in the SGC shutting down due to an unchecked IllegalStateException during task data pool exhaustion. (SSSVN-395)

Fixed an issue where segments could arrive out of order if sent while the task pool was exhausted. (SSSVN-397)

Fixed NullPointerExceptions that could sometimes be thrown when requesting SNMP counters for OGT and DPC info. (SSSVN-398)

Fixed an issue where under certain conditions all worker threads could be hanging around waiting to be allowed to send, resulting in no worker threads available to process incoming messages (such as heartbeats required to maintain connectivity). (SSSVN-400)

Fixed an issue where segments received split across multiple stream IDs could result in the SGC exiting due to an uncaught exception. (SSSVN-406)

Fixed an issue which could prevent the SGC from successfully restarting. (SSSVN-364)

Fixed an issue with SCCP decoding when the Data part comes after the Optional Parameters part in the received message. (SSSVN-372)

Fixed a NullPointerException caused by insufficient tasks being available in the task pool. (SSSVN-201)

Fixed AssertionError: Empty buffer: Cannot decode message when transferring messages internally from one SGC to another. (SSSVN-352)

Fixed an issue which could cause the SGC not to restart automatically following a graceful shutdown caused by an uncaught exception. (SSSVN-249)

Fixed an issue causing the CLI and MBeans to display stale status information for asinfo, associationinfo, dpcinfo, and pcinfo on failed SGCs. (SSSVN-202)

Fixed a race condition that could cause the SGC to attempt to register the same MBean twice. (SSSVN-159)

Handled the failure to deliver NOTICE to TCAP stack in an edge case. (SSSVN-165)

The uncaught exception handler is now initialized earlier. (SSSVN-248)

The example hazelcast.xml.sample file now matches the in-jar default hazelcast configuration. (SSSVN-235)

Disabled the hazelcast SystemLogService by default as it has a small memory leak. (SSSVN-192)

Fixed a NullPointerException in OgtInfoHelper.compare, which resulted in display-info-ogtinfo, snmpwalk and retrieving snmp stats to fail under certain conditions. (SSSVN-363)

Fixed the SGC system OID to be the correct OID. (SSSVN-371)

The SGC no longer spams stdout (or the startup log) with state info. (SSSVN-146)

Outbound global title translation errors are now logged at WARN level. (SSSVN-43)

Added generate-report.sh script for easy log gathering. (SSSVN-191)

The default Hazelcast configuration has been changed so that the SGC now detects cluster member failure more quickly. (SSSVN-295/SSSVN-280)

Changed the default values of ocss7.trdpCapacity , ocss7.schNodeListSize, ocss7.taskdpCapacity, ocss7.wgQueuesSize, and ocss7.senderQueueSize; this provides a more coherent default configuration. (SSSVN-274)

Fixed an issue which caused the SGC not to raise alarms for certain configuration items (ASes, associations, etc) if it was unable to activate them after starting up. (SSSVN-129)

Fixed a issue which would cause the SGC to throw an AssertionError and exit abnormally after receiving a TC-BEGIN which could not be handled because of insufficient resources. (SSSVN-302)

Fixed an issue which could cause the SGC to crash under certain M3UA failure conditions. (SSSVN-196)

Fixed an issue which could cause DPC configuration changes for segmentation not to be applied when changed. (SSSVN-211)

An error is now returned to the CGIN user when attempting to initiate a dialog to a global title longer than the supported maximum. (SSSVN-53)

Fixed a SGC crash that could occur when unregistering alarms. (SSSVN-255)

display-info-associationinfo no longer intermittently displays out of date information following disabling of an active association. (SSSVN-285/SSSVN-279)

Fixed an issue with the CLI which caused it to report unknown command errors when the connection to the SGC was lost. (SSSVN-99)

SCCP XUDT support extended to include reassembly of incoming segments, segmentation of outgoing messages, and configuration of segmentation parameters and preferred message type per destination Point Code. (SSSVN-79)

Added SGC startup option --seed, which can be used to provide an alternative entropy source to the SGC’s encryption facilities. (SSSVN-116)

When deactivating an M3UA connection to a peer the SGC now sends ASP-INACTIVE and ASP_DOWN rather than simply using the SCTP Shutdown procedure. (SSSVN-123)

SGC local port configuration validation prevents the use of ports below 1,024. (SSSVN-124)

The SGC will now restart correctly if terminated by an uncaught Exception or Error. (SSSVN-157)

Fixed an issue causing SCTP association down alarms to be removed and replaced periodically while the association remained down. (SSSVN-120)

Fixed MIB syntax errors which prevented some SNMP clients from accepting the OCSS7 MIB. (SSSVN-1)

Fixed timestamps sent by SNMP, which now send the day of the month, but used to send the day of the year. (SSSVN-2)

Corrected the spelling of CLI configuration parameter table.format.maxCellContentLength. (SSSVN-51)

Input is now validated when changing boolean configuration attributes. (SSSVN-61)

Fixed an SCCP handling bug which could cause the SGC to exit if an unrecognised optional parameter was received. (SSSVN-126)

Fixed an issue allowing overly long GTs to be placed in inbound-gtt configuration. (SSSVN-152)

Fixed in issue allowing overly long GTs to be placed in replace-gt configuration. (SSSVN-54)

The SGC’s SNMP MIB description strings have been improved. (SSSVN-58)

If both global title and point code are absent on an incoming SCCP message, set the SCCP OPC to that of the L3MTP OPC, in order that we’re able to route responses to this message. (SSSVN-104)

Support A, D and E sccp address global title digits (defined as spare in Q.713). (SSSVN-42)

Inbound GTT rules may omit the SSN, allowing the CdPA SSN to be retained. (SSSVN-4)

Inbound GTT rules with duplicate NAI/NP/TT/digits can no longer be created. (SSSVN-5)

Fixed incorrect outbound GTT for messages with: GT present, PC present, SSN absent. (SSSVN-11)