The Enable Discovery at Startup checkbox on the NerveCenter Administrator's Server page was removed.
For information about doing a ping sweep of your network, refer to the IPSweep behavior model discussed in the NerveCenter Behavior Models Cookbook.
The NerveCenter Server now has read-execute permissions in the model directory and its subdirectories, and does not terminate because of permissions issues.
NerveCenter now tracks SNMP requests dropped and the SNMP errors counter is now fixed.
NerveCenter alarm instances did not always transition when expected
In previous versions of NerveCenter, poll-based alarm transitions away from a non-Ground state could take place at any time: from whenever the alarm instance first reached the state, up to the full poll period for the poll causing a transition. In most circumstances, it is preferable for a poll to take place right away when an alarm instance reaches a state it could be transitioned out of by that poll. With this version of NerveCenter, all relevant polls take place as soon as an alarm instance could be affected by those polls. If a delay is desired, the correct procedure is to use a FireTrigger alarm action to force a delay. Poll rate settings only apply to looping transitions on the same state.
Using localtime() in NerveCenter Perl subroutines made the NerveCenter Server unresponsive
Using localtime() calls in NerveCenter Perl subroutines and trap masks no longer makes the NerveCenter Server unresponsive.
The small memory leak in the NerveCenter Server when you turned polls (that are in use by an alarm) on and then off again has been corrected.
When NerveCenter is installed co-resident with OpenView (OV), in addition to normal informational traps that OV forwards to NerveCenter, OV also sends traps when map and other changes occur. In addition to increasing its trap processing efficiency in general, NerveCenter implemented the -drop17traps command line flag to ignore those spurious OV traps.
NerveCenter's queues might back up with heavy socket traffic.
During periods of really high CPU utilization, NerveCenter did not flush out its pending lists, timeouts, and so on. As the list grew, NerveCenter slowed down because it took more time to search. This problem has been corrected in version 3.7.
Varbind 16 of an InformOV in NerveCenter v3.7 now contains the original trap's data (TrapTime, Generic Trap Number, Specific Trap Number and Enterprise OID).
NerveCenter now handles properly an SNMPSet value for the ciscoPingMib's ciscoPingAddress attribute.
When creating poll conditions using ccmHistoryEventEntry.ccm, the following error occurred:
NC Poll Manager 3002: CPollManagerWnd::OnPollOnOff, PreCompile of PollEvent with Poll Id 33 failed.
NerveCenter did not convert enumerated values to text from integer
NerveCenter now converts enumerated values to text from integer in log files. Text information can be used to generate informative summary messages in Network Management Platforms, such as NetCool.
NerveCenter OpenView Platform Adapter (ovpa) did not create PC.dat files
NerveCenter now creates a PC.dat file on HP-UX.
(PC.dat files are used to upload parenting configuration information into the NerveCenter node list using the ovpa -pc hostname command.)
When the same poll was used as a trigger source multiple times within a behavior model, the first poll occurred at the beginning of that poll's polling interval. The poll was scheduled at the end of the interval if used later in the state diagram. This problem has been fixed.
Triggers were not firing on routers with large numbers of interfaces
Some alarms were not transitioning for nodes NerveCenter was polling. The reason for this was that NerveCenter was firing too many SNMP requests too fast, and the SNMP agents on the nodes being polled were dropping some requests. NerveCenter has changed its polling to enable nodes to have time to respond to polls.
Manage node with port other than 161 and poll uses port 161 anyway
For an SNMP poll, if the node had multiple interfaces and the SNMP port number for that node was not 161, the NerveCenter Server fired SNMP_TIMEOUT for that node. This problem has been fixed.
The NerveCenter LogToFile alarm action no longer truncates when del% is zero.
The NerveCenter ping and SNMP buffer size is now an appropriate size: 256KB for Windows and Solaris; 128KB for HP-UX.
NerveCenter performance, with respect to the way in which it calls host names, has been significantly improved.
NerveCenter can now ICMP ping the host on which the NerveCenter Server runs.
Polls were not scheduled for tabular MIB entries that have a zeroth entry. This problem has been fixed.
NerveCenter Trapper did not restart automatically if it was killed.
The NerveCenter Server now restarts Trapper automatically within one minute of Trapper terminating.
Large numbers of instances in subobject scope alarms degraded NerveCenter performance
NerveCenter was sending SNMP requests, receiving replies, and firing triggers into its internal alarm manager, which was processing this data inefficiently. An internal NerveCenter data structure change has been made and this issue is no resolved.
On HP-UX systems, the HTML version of the NerveCenter Release Notes was not installed and the NerveCenter HTML documentation's Search feature was disabled, unless you mounted the NerveCenter CD with the -o cdcase switch. This problem has been corrected.