README IBM Tivoli Network Manager IP Edition 3.7 Fix Pack 4 README
(Build 3.7.0.23)


Introduction
============
This README contains information on issues fixed in IBM Tivoli Network Manager IP Edition 3.7 fix pack 4.

For more information about V3.7 please refer to the latest version of the Release Notes.

This build should be used with Solaris 9 or 10 for SPARC, Redhat Enterprise Linux 3 or 4, SUSE Enterprise Linux 9 or 10, AIX 5.3 and Windows 2003 Server operating systems.


Prerequisites and Installation

==============================

IBM Tivoli Network Manager IP Edition 3.7 fix pack 4 is available as separate installs for Solaris, AIX, Windows and Linux.

For more information on prerequisites and minimum recommended system requirements, refer to the IBM Tivoli Network Manager IP Edition 3.7 Release Notes and Installation and Deployment Guide.

All Network Management processes running on the server to which the FixPack is being applied must be STOPPED.

For ITNM Entry Edition, or installations utilizing the Reporting and Administration Toolset, on Unix the processes may be stopped in the following manner:

1. Source "[$NCHOME]/bin/env.sh"

2. Stop Network Manager solution using "itnm_stop" prior to installing FP. This will stop all Network Manager, OMNIBUS, Netcool Security Manager, Netcool GUI Foundation and License Manager processes.

3. Stop the mysql process via the mysqladmin shutdown command.

For ITNM IP Edition on Unix the processes may be stopped in the following manner:

1. Use the ncp stop script:
$NCHOME/precision/scripts/ncp stop

2. Stop the GUI Foundation server:
ngf_server stop

3. Stop the mysql process via the mysqladmin shutdown command.

For both ITNM Entry Edition and ITNM IP Edition on Windows the processes may be stopped in the following manner:
Use "Services" to stop processes in following order, ncp_ctrl, NCO Multi-threaded TRAPD Probe, Netcool GUI Foundation Server", "Netcool Security Manager Server", "Netcool OMNIBUS Object Server", "NCO Flex License Manager", "Netcool MySQL".

Verify all processes stopped

* On Unix use "itnm_status" and "ps -ef" to ensure all processes are stopped.

* On Windows use "Task Manager".

Specifically, on Unix look for the "rvd" or "rva" processes or any process starting with "ncp". On Windows, look for the "rvd.exe" or "rva.exe" processes or any process starting with ncp. If any of them are still running they will need to be stopped manually:

UNIX "kill [#pid#]" or Windows "Task Manager" select and End Process.

UNIX
----

Enter the following command to install this fix interactively. The placeholder should be identified with the string identifying your platform.

$NCHOME/install/ncisetup -install 3.7.0.23-NetworkManagerIP--FP0004.jar

Enter the following command to install this fix non-interactively. The placeholder should be identified with the string identifying your platform.

$NCHOME/install/ncisetup -install 3.7.0.23-NetworkManagerIP--FP0004.jar -silent

On AIX, the installer is bundled with a wrapper script called "INSTALL" that increases the data size limit for the shell before installing the fix pack. You must set the $NCHOME environment variable before running the "INSTALL" script. Alternatively, type "limit datasize unlimited" (csh/tcsh) or "ulimit -d unlimited" (sh/ksh/bash) at your shell prompt and then install as above.
Additionally, the way to allow the installer to use up to 2GB of memory is to type this before starting the ITNM FixPack installation: "ldedit -b maxdata:0x80000000 $NCHOME/install/platform/aix5/bin/ncisetup"
You may also have to set the env prior to running $NCHOME/install/ncisetup -install 3.7.0.23-NetworkManagerIP--FP0004.jar -silent

Windows
-------

Unzip 3.7.0.23-NetworkManagerIP-win-FP0004.zip and double-click setup.exe to install this fix.

ALL
---
Once the above steps are completed, please make sure to apply the steps described below for the tickets:
- 53312
- 54838
- 55418
- 57648


Known Issues with V3.7 fix pack 4 (Build 3.7.0.23)
==================================================
52195 Disco Admin GUI: Saving discovery configuration and other GUI settings
IZ07978 54471 SNMP HELPER CAN MIS-FORMAT OCTET STRINGS WITH NON-PRINTABLE CHARACTERS
IZ09413 54832 DIRECT ROUTER TO ROUTER LINKS ON JUNIPER DEVICES SHOULD BE ON PHYSICAL INTERFACES
56360 Error with ncp_poller and DB2 9.1

Applying Fix Pack on Windows Network Manager server removes the "Automatic" starting of ncp_ctrl process in "Services", You may wish to set it back to "Automatic".
The rvd or rvd.exe process does not always exit when Network Manager solution is completely stopped.

Summary of Fixes in V3.7 fix pack 4 (Build 3.7.0.23)
====================================================
53103 IpForwardingTable agent and support partial matching
-----------------------------------------------------------------------------------------------------------
Performance of the IpForwardingTable agent improved. New configuration option added to restrict route download (see .agnt file)

IZ10117 54956 PREVIOUS SNMP QUERY IS EXECUTED AUTOMATICALLY WHEN WE REOPEN THE MIB BROWSER.
-----------------------------------------------------------------------------------------------------------
Problem was addressed, when navigating back to MIB browser screen is reset and no previous query gets executed.

IZ12239 55099 JPS404 ERRORS WHEN CREATING THE MIB PICKER TREE IN THE GUI
-----------------------------------------------------------------------------------------------------------
Page printing error messages were missing. Instead, when data in ncmib database was corrupted or missing, 404 error was printed. Proper error page was added and log messages in log file were enhanced.

IZ12942 55185 IN A MULTI-DOMAIN, THE MIB BROWSER USES WRONG SETTINGS
-----------------------------------------------------------------------------------------------------------
Additional configuration file was introduced which stores data specific to each of domains. MIB browser will select appropriate information about domain basing on user's selection.

IZ14207 55282 EXTEND ADJUSTEDIPLAYER TO HANDLE CISCO SUBINTERFACE LINKS
-----------------------------------------------------------------------------------------------------------
Improved AdjustedIPLayer.stch support for Port Channel interfaces and Cisco-Cisco subinterface links

IZ14329 55313 DEFAULT.XML IN AUTOPROVISION DIRECTORY DOESN'T MOVE OR LOG
-----------------------------------------------------------------------------------------------------------
The system has been modified to produce warning messages in the Topoviz log file when either of these scenarios occur. In the case that any provision file exists along with its ".processed" counterpart, the file will not be processed

IZ14824 55405 PROVIDE SPECIFIC FUNCTION TO ITNM IP 3.7 THAT HAD BEEN AVAILABLE IN 3.6. THE VIEW FRAME ON THE BOTTOM OF THE NETWORK & HOPE VIEW.
-----------------------------------------------------------------------------------------------------------
Simple replacement for Connectivity Information functionality was added. To enable it follow steps below:
1) Add new "Show Connectivity Information" tool. Create $NCHOME/etc/percision/tools/showConnectivityInformation.xml file and add there following lines:
2) Edit $NCHOME/etc/percision/menus/ncp_topoviz_device_menu.xml and add created tool to Network View menu using following definition:
3) Restard NGF server, clear Java browser cache.
After above steps, right click on particular device and select "Show Connectivity Information" tool. Separate window will appear presenting connectivity details.
By default all interface types are printed. To exclude some types of interfaces you can define which interfaces should be included and which excluded. To do this edit $NCHOME/etc/precision/topoviz.properties file and add one of two lines: topoviz.connectivity.include=IF_TYPE1, IF_TYPE2, ... topoviz.connectivity.exclude=IF_TYPE1, IF_TYPE2, ...
Types of interfaces should be comma separated numbers. To see all available numbers run following query on NCIM database:
SELECT * FROM ncim.enumerations e where enumGroup = 'ifType';

IZ14944 55418 STATUS ICONS SLOW
-----------------------------------------------------------------------------------------------------------
Performance improved by implementation of new caching layer. This requires two extra tables to be added to the database. Please follow the appropriate instructions for adding these tables to your database. These commands can be executed on a running database.
If these commands are not executed then warnings will be shown in $NCHOME/log/precision/ncp_topoviz.0.log and the system will fall-back to it's uncached implementation and work as before.
MySQL:
$NCHOME/platform//mysql5.0/bin/mysql ncpgui -u root < $PRECISION_HOME/scripts/sql/mysql/updatePrecisionGUIDb.sql
should be set as follows:
For Solaris : solaris2
For Linux : linux2x86
For AIX : aix5
For Windows : win32
Additionally, on Windows replace $NCHOME with %NCHOME% and $PRECISION_HOME with %PRECISION_HOME%.
DB2:
Copy the contents of $PRECISION_HOME/scripts/sql/db2 to your DB2 server and run the following command from the directory you copied them to.
Unix: updatePrecisionGUIDb.sh
Windows: updatePrecisionGUIDb.bat
Where , and are set to your database name, username and password as appropriate.
Oracle:
Copy the contents of $PRECISION_HOME/scripts/sql/oracle to your Oracle server and run the following command from the directory you copied them to.
sqlplus / < updatePrecisionGUIDb.sql
Where and are your username and password as appropriate.

IZ15160 55436 DISCOVERY DOESN'T GET INITIATED WHEN CONFIGURED FOR EMS INTEGRATION (COLLECTORS) AND NAT ENVIRONMENT
-----------------------------------------------------------------------------------------------------------
Fixed issue in NATAgentRetProcessing.stch which prevented Collector based discovery when NAT is enabled.

IZ15287 55452 Disco Admin GUI: MEMORY LEAK IN PRECISION 3.6
-----------------------------------------------------------------------------------------------------------
Memory leak in Discovery GUI was eliminated by changing the way how memory resources were utilized.

IZ15447 55485 FAILOVER IS TRIGGERED UPON REDISCOVERY DUE TO FAILURE OF THE VIRTUAL DOMAIN PROCESS TO GENERATE HEALTH CHECK EVENTS.
-----------------------------------------------------------------------------------------------------------
The OQL messages to update the config.failover tables of the listening processes are now sent in a separate thread, so as to prevent the main functionality of the virtual domain process from being blocked.
The gateway also fast-tracks health check messages from the Object Server to the virtual domain process. The eventMap is no longer used for these events.

IZ15515 55509 ITNM 3.7 FAILS TO POLL DEVICES USING SNMPV3 AUTHTYPE=SHA1
-----------------------------------------------------------------------------------------------------------
Incorrect internal handling of the configuration led to the wrong key size being used for SHA key generation.

IZ15546 55510 PARTIAL REDISCOVERY IS NOT WORKING AS EXPECTED ,WHEN SELECTED FROM REDISCOVER NODE(S) TOOL.
-----------------------------------------------------------------------------------------------------------
Partial rediscovery did not work properly when initiated on a node from Network View. The reason for that was that domain name was used instead of IP address of the devide to be discovered. Discovery process requires from GUI passing IP address / subnect address in order to initiate partial rediscovery. Domain name is now resolved and IP address passed to discovery process which addresses the issue.

IZ15752 55544 MPLS CLOUD ALWAYS APPEARS EXPANDED. SAVING CHANGES TO MPLS NETWORK VIEWS DOES NOT SEEM TO BE WORKING.
-----------------------------------------------------------------------------------------------------------
Network View code was fixed to handle save of MPLS clouds.

IZ15986 55582 SOME PERL AGENTS ARE FAR TOO SLOW; SPECIFICALLY REQUIRE CISCO VLAN MODELLING AND CISCO VTP TO BE COMPILED
-----------------------------------------------------------------------------------------------------------
Because Perl Entity agent is single threaded, it causes very large delays in discovery. so CiscoRoutedVLANInterface, CiscoVTP agent rewritten and compiled in C++

IZ16709 55765 NCP_DH_PING FAILS IF M_BROADCAST = 1
-----------------------------------------------------------------------------------------------------------
Fix for problem with ping helper when broadcast or multicast pinging is configured as acceptable.

IZ16963 55837 DISCOVERY DETAILS TAB, PING FINDER PROGRESS, CURRENT ADDRESS ALWAYS SHOWS 'NULL'. 3.7 FP2.
-----------------------------------------------------------------------------------------------------------
Fix to issue where the ping finder was not updating the pingFinder.status table with it;'s progress in subnet pings. This hence prevented the GUI from displaying this information.

IZ17015 55845 COLLECTORDETAILS AGENT SHOULD HAVE DNS LOOKUP FEATURE
-----------------------------------------------------------------------------------------------------------
Discover will now perform DNS resolution for Collector discovered entities when the Collector data doesn't contan an entity name. Alcatel 5620 SAM Collector modified not to send a name.

IZ17126 55897 UNABLE TO SEE MIBS FOR DEVICES LIKE CISCO AND JUNIPER IN THE MIBBROWSER
-----------------------------------------------------------------------------------------------------------
Fix to ASN1 ordering bug. This bug could lead to incomplete oid definitions returned from the mib browser.

IZ17311 55910 AMOS CONNECTED SUPPRESSION RULE CAN CAUSE CIRCULAR SUPPRESSION
-----------------------------------------------------------------------------------------------------------
All users should update two of their AMOS rules files, namely a new version of the file SuppressConnectedAlerts.rule should be obtained AND the newly created rules file SuppressContainedAlerts.rule should be obtained too and place in the rca_rules directory with the other rules files.

IZ17397 55932 MEMORY LEAK IN NCP_POLLER WHEN SNMP POLLS ARE ENABLED
-----------------------------------------------------------------------------------------------------------
Memory leaks in libNcpSnmpThreshold.so have been addressed.

55969 Too many file descriptors are being opened by eval($MEMUSAGE)
-----------------------------------------------------------------------------------------------------------
The linux version of the code opened the socket but didn't close it. This has now been addressed

IZ17665 55971 DISCOVERING JUNIPER L2 VPN VIA JUNIPER SNMP AGENT
-----------------------------------------------------------------------------------------------------------
Support for layer 2 VPN data implemented

IZ17668 55973 PERL API FROM VERSION 3.6 ONWARDS CAN NO LONGER LISTEN FOR UPDATES ON THE TOPOLOGY NOTIFICATION SUBJECT
-----------------------------------------------------------------------------------------------------------
A code change was made to fix this bug.

IZ17669 55976 UNABLE TO EDIT/VIEW EXISTING/NEW POLL DEFINITIONS ON DISTRIBUTED INSTALLATION HAVING ORACLE AS NCIM DATABASE
-----------------------------------------------------------------------------------------------------------
There was a problem with data type conversion on Oracle - Java JDBC driver. Problem was addressed by modifying code to use different type for reading variables from ncomitor database.

IZ17722 55983 EVENT SEVERITY VALUES CANNOT BE UPDATED FROM THE POLL DEFINITION EDITOR IN THE MONITOR ADMIN GUI.
-----------------------------------------------------------------------------------------------------------
Changes done in Poll Definition Editor od Monitor Administration GUI caused error message and were not persisted in the database. Problem affected only users using DB2 as the NCIM database engine.
DB2 JDBC driver does not allow more than 1 concurrent updatable ResultSet per databse connection. Changes were implemented in order to have only one updatable ResultSet at the same time.

IZ17814 55999 THE POLLER PROCESS CAN OVERWRITE THE SNMP CONFIGURATION STORED IN THE DATABASE BY THE SNMP HELPER
-----------------------------------------------------------------------------------------------------------
Under normal circumstances, the poller relies on the SNMP helper to write the configured SNMP acess details [SnmpStackSecurtiyInfo config information] to the ncpoller database.
The poller will only write the configured SNMP access details to the ncpoller database if it is running in the backup domain in failover configuration. In this situation, it is expected to be using a distinct database, as the poller DB is typically set up locally, and a failover architecture, by its nature, is distributed across at least 2 different servers. As the SNMP helepr is not used in the backup domain, the poller must read in the configured information.

56058 Poor performance in ncp_model due to compound primary key
-----------------------------------------------------------------------------------------------------------
The old composite primary key was replaced by a new primary key

IZ18625 56119 LARGE MEMORY CONSUMPTION WHEN ARE TOO MUCH DATA FROM DISCOVERY
-----------------------------------------------------------------------------------------------------------
Problem was addressed by optimizing memory usage in ITNM GUI and also option to disable automatic discovery status update functionality.
In order to use this settings, edit or create $NCHOME/etc/precision/discovery_gui.properties file and add 'discovery_gui.status.allowAutoUpdate' and set it to 'false' in order to disable automatic discovery status update, example:
discovery_gui.status.allowAutoUpdate=false
By default, automatic status updates are enabled.

IZ18948 56182 MANY LINKS HAVE FDB DATA NOT GETTING POPULATED
-----------------------------------------------------------------------------------------------------------
The optional Analyse & Prune stitchers have been enhanced to improve connectivity on networks with incomplete switch forwarding database information.

56218 Improvements to handling out-of-memory errors in C++ processes
-----------------------------------------------------------------------------------------------------------
Whenever program runs out of memory, it was not handled properly. i.e. the exit sequence and error message was incorrect. Which was not clear to normal user. Now whenever program runs out of space it wont generate any core and exits from execution at that point by logging message.

IZ19476 56331 DNS HELPER WAITS FOR FULL TIMEOUT EVEN THOUGH IT HAS RECEIVED A REPLY AND COULD RETURN IMMEDIATELY
-----------------------------------------------------------------------------------------------------------
Basically what happens is that the DirectDNS method sends a dns packet to a dns server requesting a name for an ip or vice versa If dns server has a problem with the packet, it responds with an error immediately send the request back to the agent saying we don't know. (earlier it wait for timeout)

56351 IMPLEMENT SNMP EXPONENTIAL BACKOFF FOR SNMPGETS SAME AS FOR SNMPWALKS AS IN PREVIOUS VERSIONS
-----------------------------------------------------------------------------------------------------------
Implemented an exponential back-off for devices we have previously got an SNMP response from. The number of extra retries is specified in the snmpHelper.configuration table, which can be modified in DiscoSnmpHelperSchema.cfg. The new behaviour is basically: if a device doesn't respond within the standard timeout/retries, we begin the exponential back-off - we keep retrying the request, but double the timeout each time. If at the end of this time, we still don't get a response, then we time out as normal, but if we get a response then we resume the operation. This behaviour previously only applied to table traversals by the SNMP helper. It now applies to SNMP Get requests as well. It doesn't currently affect the ncp_poller process.

56456 When Network View is edited, the changes takes place before it is saved, but when the save button is clicked, the changes doesn't takes place and it reverts back.
-----------------------------------------------------------------------------------------------------------
Cache cleaner now just logs a message telling the user to apply the database changes.

56525 ncp_f_amos exits with parse error in SuppressConnectedAlerts.rule
-----------------------------------------------------------------------------------------------------------
The typo error in the rules file has been removed.

IZ20193 56539 DEVICES ARE NAMED WITH IPADDRESS RATHER THAN SYSNAME
-----------------------------------------------------------------------------------------------------------
The problem has been resolved by updating the stitchers.

IZ20552 56637 BUILDFINALENTITYTABLE STITCHER FAILS TO BUILD A RECORD UPON RECEIVING RECORD WITH IFINDEX = 0 VALUE.
-----------------------------------------------------------------------------------------------------------
The error message is logged for every record which has ifIndex value 0. If customer has large number of records with if index 0 then logg will be populated with huge amount of error messages. In this fix the debug level is raised from 0 to 4 so that these error messages populate only if you set debug level to 4

IZ20816 56712 PING FAIL EVENTS POSSIBLE DURING REDISCOVERY
-----------------------------------------------------------------------------------------------------------
The source of these packets is no longer identified as the failing (unreachable) entity.

IZ20875 56729 NCP_POLLER FAILS WHEN USING SNMPV3
-----------------------------------------------------------------------------------------------------------
An internal OctetString type is used where appropriate.
Problems selecting USM users and discovering access for SNMPv3 devices have been resolved.

IZ20876 56731 NAME RESOLUTION IN DUAL DOMAIN DNS SERVER APPENDING TWO DOMAIN
-----------------------------------------------------------------------------------------------------------
The DNS helper is currently written to manage only a single name domain per DNS method. This can cause an issue with name services which respond with names for multiple name domains. The effect is that some entities may contain multiple name domains. e.g. router1.namedomain1.com.namedomain2.com

IZ20899 56750 WHENEVER PROGRAM RUNS OUT OF MEMORY, IT WAS NOT HANDLED PROPERLY.
-----------------------------------------------------------------------------------------------------------
The following changes have been made:
1) In RivErrorCodeStrings.cc, RivErrInsufficientMemory should be changed from type E_RESFatal to E_RESExit
2) Also in RivErrorCodeStrings.cc, change the error message for RivErrInsufficientMemory to this: Out-of-memory - This process has run out of memory or exceeded the operating system memory limit for a 32 bit application - You must take action to reduce the amount of data this application is having to process
3) In CRivError.cc, change EMERGENCY_MEMORY_SIZE from 2048 to 32768 (2K to 32K). Hopefully this will be enough to get the error message printed, even in heavily multi-threaded processes.

IZ20999 56778 JUNIPERMPLSTELNET AGENT FAILS DURING RETRIVAL OF DATA FROM TREE LIST DUE TO UNDEFINED KEY TYPE
-----------------------------------------------------------------------------------------------------------
Now the variable is being checked before the message is printed, preventing the failure

IZ21276 56863 ITNMV3.7-NCP_MODEL FAILS DUE TO STACK OVERFLOW, "CMDLLINGERTIME::FIXLINGERINGCONTAINS"
-----------------------------------------------------------------------------------------------------------
The fix was to change the OQL query. This will not result in a deeply nested filter tree and will therefore not cause a stack overflow.

56888 AMOS can get stuck in its daemon-killing function, especially on AIX
-----------------------------------------------------------------------------------------------------------
A fix was made to prevent Amos getting into a situation in which it can get stuck and needlessly use a lot of CPU for a while and which potentially can result in memory leaking. This only happened under certain circumstances of event/topology combinations, so luckily was fairly rare it seems.

IZ21383 56915 IMPROVEMENT TO THE LAYER-2 CONNECTIVITY
-----------------------------------------------------------------------------------------------------------
network is made of HP switches and the CDP protocol is holding remote nbrs ifIndex instead of ifname that is confusing CDP layer stitcher

IZ21471 56960 ITNMV3.7 - FAILOVER NOT WORKING.
-----------------------------------------------------------------------------------------------------------
ncp_ctrl sets the serviceState of hung processes in services.inTray to 'Sick' = 5.
In a failover system, the primary ncp_virtualdomain will not initiate fail over until it has had a socket connection to the backup domain.
A health check failure event will still be raised if a problem is detected.
In summary:
1) The primary domain will not fail over if it has never established a connection to the backup. This allows the primary to be run in isolation.
2) The backup domain can be run in isolation - it will become primary after the configured failover time (m_FailoverTime in VirtualDomainSchema.cfg). This handles the case where a primary server has completely gone down, and the backup is restarted.
3) The primary and backup domains can be started in any order. It is still recommended that the backup domain not be started until the primary virtual domain process is up and running, as this will avoid the possibility of temporararily having both domains simultaneously monitoring the system, and generating the associated network traffic.
4) The configured m_FailoverTime must be at least 3 times the m_HealthCheckPeriod for correct initialisation.

IZ21931 57037 NCP_POLLER PROCESS SOMETIMES HANGS WHEN RUNNING ICMP POLLS
-----------------------------------------------------------------------------------------------------------
Deadlock is now avoided in SnmpPoller module.

IZ22005 57056 NCP POLLER DOESN'T START IN FAILOVER CONFIG
-----------------------------------------------------------------------------------------------------------
SQL has been updated to query the primary domain.

IZ22021 57072 ITNMV3.7 - POLLER MOVES INTO DEADLOCK STATE.
-----------------------------------------------------------------------------------------------------------
The problem is due to a bug in the SNMP Get aggregation code, which attempts to reduce the number of SNMP requests emitted by the ncp_poller process by 'packing' PDUs. The problem is avoided by disabling this functionality.

57128 Disco Admin GUI: scratchTopology column names need to be updated.
-----------------------------------------------------------------------------------------------------------
The stitchers and discovery schema have been changed. The scratchTopology.entityByName table now has the same fields as the master.entityByName table in model. The stitchers have been altered to use the table in this form and to use the new rule to take the records in this form and send them to model. This fix addresses modifications required by ITNM GUI to use modified schema properly.

57197 Networkview - when user creates filtered view with an invalid SQL filter, DB locks on few tables does not get released, causing exceptions thrown to user.
-----------------------------------------------------------------------------------------------------------
While creating view if user manually enters a filter string with invalid SQL syntax (ex: invalid column, etc), logs give series of exceptions with messages like 'table was not locked' and GUI showed Http 500 error. Problem was addressed by rolling back transaction, unlocking tables and presenting user friendly error message.

57236 Minor bug in the month of timestamp for filename created by GetDiscoCache.pl
-----------------------------------------------------------------------------------------------------------
The month of the timestamp is out by a month as gmtime returns month as 0 to 11. Changing it to 1..12

IZ22383 57264 SNMPHELPER CAUSES LONG DISCOVRIES UPON FAILING TO GET RESPONSE ON VLAN REQUESTS AND AWAITS THE TIMEOUT PERIOD(20MINS DEFAULT)
-----------------------------------------------------------------------------------------------------------
Efficiency of CiscoStp improved for devices lacking community string indexing support

IZ22407 57265 MULTILINK PPP LINKS CONTAINAMENT BUILDING IS INCORRECT WHEN IFSTACKTABLE AGENT DISCOVERS 2 OR MORE LOWER NEIGHBOURS
-----------------------------------------------------------------------------------------------------------
This is now fixed, changes made as-per Nick Ho's modifications

57304 ncp_poller fails on exit when connected to DB2
-----------------------------------------------------------------------------------------------------------
The problem with ncp_poller coring on exit with DB2 has been fixed

IZ22831 57344 SNMPPOLLER FAILS WHEN A FLOOD OF SMALL SNMP REQUESTS HAPPEN IN A SHORT PERIOD OF TIME
-----------------------------------------------------------------------------------------------------------
Data required to resolve alert descriptions is now collected as part of the primary poll.

IZ23027 57437 EMPTY OQL STATEMENT, SYNTAX ERROR, WARNING MESSAGES IN DISCO LOG.
-----------------------------------------------------------------------------------------------------------
The warning messages were being produced due to an empty string being used for an oql insert. We added a check so that the insert is not attempted if there is an empty string which prevents the warnings.

57460 improve default configuration to improve SNMP performance and reliability
-----------------------------------------------------------------------------------------------------------
Updated ncp_disco.war to use 2 as the default value for SNMP retries, and 3000 for timeout
Updated the default SnmpStackSecurityInfo.cfg to include m_NumRetries and m_TimeOut in the default insert into snmpStack.verSecurityTable.
Also removed the SNMPv3 config which is not used anyway, as the entry is configured for SNMPv1.
Updated DiscoSnmpHelperSchema.cfg to specify 0 for m_ExponentialRetries by default.

IZ23239 57544 DOWNLOADING DISCOVERY STATUS DATA CAN BE EXCESSIVE
-----------------------------------------------------------------------------------------------------------
The OQL language was augmented with the addition of a COUNT(*) function and the disco configuration GUI changed to use this new function when totalling up the number of rows in the instrumentation tables.

IZ23324 57648 INFERRED CES ARE MISSING FROM THE NCIM IPENDPOINT TABLE
-----------------------------------------------------------------------------------------------------------
The CE IP address now appears in the NCIM ipEndPoint table. This requires extension of the ipEndPoint.addressSpace field, which can be done using the updatePrecisionIPDb scripts, found in $NCHOME/precision/scripts/sql//. Instructions for use can be found at the top of the appropriate script.

IZ23334 57649 MODEL TAKING TOO LONG TO START WHEN NCIM IS ON ORACLE
-----------------------------------------------------------------------------------------------------------
ncp_model was taking a long time to download enumerations from NCIM at startup. In particular, there are tens of thousands of enumerations for the 'ASN' group. This information is availbale in disco, and now populates the record for each appropriate entity before it passed to model, so the enumerations are no longer required by model itself.



Known Issues with V3.7 fix pack 3 (Build 3.7.0.20)
==================================================
52195 Disco Admin GUI: Saving discovery configuration and other GUI settings
IZ07978 54471 SNMP HELPER CAN MIS-FORMAT OCTET STRINGS WITH NON-PRINTABLE CHARACTERS
IZ09413 54832 DIRECT ROUTER TO ROUTER LINKS ON JUNIPER DEVICES SHOULD BE ON PHYSICAL INTERFACES
IZ15447 55485 FAILOVER IS TRIGGERED UPON REDISCOVERY DUE TO FAILURE OF THE VIRTUAL DOMAIN PROCESS TO GENERATE HEALTH CHECK EVENTS.

Applying Fix Pack on Windows Network Manager server removes the "Automatic" starting of ncp_ctrl process in "Services", You may wish to set it back to "Automatic".
The rvd or rvd.exe process does not always exit when Network Manager solution is completely stopped.

Summary of Fixes in V3.7 fix pack 3 (Build 3.7.0.20)
====================================================
IY97131 50509 ENHANCE CISCOSWITCHTELNET AGENT TO SUPPORT 'SHOW MAC-ADRESS-TABLE' OUTPUT OF DIFFERNT FORMAT
-----------------------------------------------------------------------------------------------------------
The CiscoSwitchTelnet agent was modified to handle the new layout of information returned from the Cisco devices.

IZ00716 52282 AMOS OVERWRITES CLEARED EVENTS WITH SEVERITY 5 WHEN LINKUP AND LINKDOWN INTERFACE EVENTS FLOWS QUICKLY
-----------------------------------------------------------------------------------------------------------
AMOS overwrites cleared events with Severity 5 when linkup and linkdown interface events flows quickly after each other.

IZ03719 53360 DATA FROM NAT TEXT FILE CAN BE IGNORED WHEN OTHER NAT AGENTS ARE ENABLED.
-----------------------------------------------------------------------------------------------------------
If you are running just the NATTextFileAgent, ensure that all other NAT agents are specifically disabled.

IZ04128 53487 NCP_DH_TELNET DOESN'T ALWAYS CORRECTLY HANDLE SERVER INITIATED CONNECTION CLOSURE
-----------------------------------------------------------------------------------------------------------
Change added to resolve hanging TCP Telnet and SSH sockets and a locked ncp_dh_telnet process.

IZ06006 53963 WHEN A MAP IS EXPORTED, THE CONNECTIONS APPEAR ON TOP OF THE DEVICES
-----------------------------------------------------------------------------------------------------------
Problem was resolved and currently exported images contain connections presented in the same way as in the network view.

IZ05666 53988 CISCOSWITCHTELNET DOESN'T PROCESS DATA IF THE HEADERS ARE UPPERCASE (EX: "MAC ADDRESS" FROM SHOW MAC-ADDRESS-TABLE)
-----------------------------------------------------------------------------------------------------------
The CiscoSwitchTelnet agent was modified to handle the new layout of information returned from the Cisco devices.

IZ06754 54277 NCP_DH_SNMP CORE DUMPS DURING DISCOVERY IF DEVICES ARE CONFIGURED FOR SNMP V3 (AIX ONLY)
-----------------------------------------------------------------------------------------------------------
Handled malloc's different behaviour on AIX in base and SNMP libraries.

54387 General DiscoAgentReturns.filter file missing from installation
-----------------------------------------------------------------------------------------------------------
Adding the missing DiscoAgentReturns.filter file.

IZ07781 54411 SUN DEVICES INSTANTIATED AS ENDNODE.AOC
-----------------------------------------------------------------------------------------------------------
The additional required filter has been added to EndNode.aoc

IZ07817 54442 THE TRANSFERRAL OF THE TOPOLOGY FROM THE PRIMARY TO THE BACKUP CAUSES THE PROBLEM HIGH CPU EVERY TIME THE 2 DOMAINS CONNECTS.
-----------------------------------------------------------------------------------------------------------
The solution was to change the way that the topology is transferred from the primary to the backup domain. The backup domain now treats the new topology as an update, as if it was received from ncp_disco. Thus, the existing topology is not deleted.
With this fix, if an existing cached topology (Store.Cache.kernel.activeModel.) is removed from primary domain before starting the Precision processes, the backup domain cache file should also be removed, to ensure consistency.

IZ08457 54528 AT RESYNCHRONISATION, THE GATEWAY SHOULD IGNORE EXISTING HEALTHCHECK EVENTS IN THE OBJECT SERVER . ONLY 'FRESH' HEALTH CHECK
-----------------------------------------------------------------------------------------------------------
We now ensure that only 'fresh' health check events, received from the Object Server over the IDUC channel, are used.
During any resync, existing health check events are ignored.

IZ08641 54568 PHASE 3 AGENTS CAN LOSE DEVICES IF PHASE 2 IS VERY SHORT
-----------------------------------------------------------------------------------------------------------
Under exceptional circumstances, the discovery could stall in phase 3, due to one or more agents that are supposed to finish in phase 3 failing to pass all devices back to the main ncp_disco process. This could happen if phase 2 finished instantly. Phase 2 will be finished instantly if the ArpCache agent is not running. The locking has now been corrected so that agents will not lose devices if a phase finishes instantly. However, note that this situation only occurs in practice when the ArpCache is not run, and it is almost always wrong to disable the ArpCache agent.

IZ08739 54590 NAMINGFROMLOOPBACKDETAILS.STCH FAILS IF TWO IPS ARE CONFIGURED ON LOOPBACK AND ONE IS 127.0.0.1
-----------------------------------------------------------------------------------------------------------
The NamingFromLoopbackDetails stitcher has been changed so that it continues searching for IP addresses on an interface if it comes across one beginning with 127.
Additionally, lo0 has been added as a priority 9 ifDescr, and lo0.0 as priority 10.
This is because on Juniper devices the loopback interfaces' ifDescrs begin with a lower case 'l' rather than upper case 'L'.

IZ08854 54668 IPTOBASENAME STITCHER SHOULD TREAT IFOPERSTATUS 7 AS MEANING AN INTERFACE IP ADDRESS IS INACTIVE.
-----------------------------------------------------------------------------------------------------------
Previously, ncp_disco would cope with duplicate IP addresses on a network if the ifOperStatus of one of the interfaces with a duplicate IP address had an ifOperStatus of 2 (down) or 6 (notPresent).
Now, it also treats interfaces with ifOperStatus 7 (lowerLayerDown) as being disabled, and hence will cope with a duplicate IP address on such an interface.

IZ08940 54706 NCP_M_TIMEDSTITCHER PROCESS TERMINATES WHEN THE IP ADDRESS STRING PASSED TO THE POLLERDOPING STITCHER RULE IS NULL
-----------------------------------------------------------------------------------------------------------
A test fix (patch) to libPollerStitcher has been provided to the customer and the same fix has been made to libPollerStitcher in ALL of versions of Precision ready for the next scheduled fixpack.

54713 SNMP helper sometimes fails to run because of permission problems when run as a non-root user
-----------------------------------------------------------------------------------------------------------

54805 CiscoMPLSSnmp agent requests mplsVpnVrfRouteNextHopAddrType column when it's not configured to do so
-----------------------------------------------------------------------------------------------------------
The TagManagedEntities stitcher will now update the managed status of existing entities if and only if it involves a transition to or from state 2. So this state - 'permanently unmanaged' - can also be thought of as 'unmanaged by disco'.
Toggling between states 0 and 1 is still performed only via the GUI.
When the managed status of a chassis is changed, the backend will set that status for all contained entities (as before), but if and only if the current status is not 2. In other words, entities with a managed status of 2 are now unaffected by changes to the maanged status of a containing entitiy.

IZ09369 54810 DISCO HANGS IF MULTIPLE COLLECTORS SUPPORT SAME DEVICE
-----------------------------------------------------------------------------------------------------------
Issue resolved which prevented multiple Collectors discovering the same device via the same IP address. It is not possible to discover a device via SNMP and multiple Collectors even if all use the same IP address to represent the device..

IZ09389 54811 COLLECTOR LAYER 3 TOPOLOGY DATA SHOULD NOT CONTRIBUTE TO THE PHYSICAL TOPOLOGY
-----------------------------------------------------------------------------------------------------------
This issue was because in CreateAndSendTopology stitcher we were also including IPlayer entities also, which is not neede. Now in this we removed that

IZ09418 54833 IMPROVEMENTS TO SONMP DISCOVERY
-----------------------------------------------------------------------------------------------------------
The SONMP discovery has been changed to:
1) Handle more ifName formats for finding local card/port.
2) Use chassis type and segment ID for finding remote card/port where possible.
3) Handle multilink trunks.

IZ09417 54834 IMPROVEMENTS TO DISCOVERY OF ALLIED TELESYN AND NORTEL SWITCHES
-----------------------------------------------------------------------------------------------------------
See diary entry dated 06/12/2007 12:59:20, as this field is too small.

54869 nco_p_ncpmonitor reports syntax error in OQL insert
-----------------------------------------------------------------------------------------------------------

IZ04103 54893 FEEDBACK STITCHER GIVES EXCESSIVE ERROR MESSAGES WHEN IT CAN'T FIND PING FINDER.
-----------------------------------------------------------------------------------------------------------
This error we were getting in particular scenario, ie when ping finder is not running and feed back stitcher is enabled. Feedback stitcher is mainly meant for ping finder.
Hence if ping finder is not there it will huge number of error message. To resolve this issue we changed feedback stitcher to check whether Ping finder is running before processing

IZ09859 54932 INCORRECT ENCODING OF LARGE ASN PACKETS
-----------------------------------------------------------------------------------------------------------
Buffer size is now correctly stored.

IZ09937 54940 MODEL IS IGNORING THE M_STATE VALUE WHEN MERGING RECORDS
-----------------------------------------------------------------------------------------------------------
Fixed by code change

54995 GetDiscoCache.pl script not responding to latency.
-----------------------------------------------------------------------------------------------------------
Fix to the GetDiscoCache.pl script to properly respond to changes in latency.

IZ10217 55013 CDP LINKS THAT RESOLVE TO SUB-INTEFACES, SHOULD BE RESOLVED TO PHYSICAL INTERFACES WHEN PROCESSED BY THE SWITCH STITCHER
-----------------------------------------------------------------------------------------------------------
Improvements made to layer 2 link resolution for cases where CDP links are to sub-interfaces. These links will now be resolved down to the physical interface.

IZ10895 55031 EXPANDING TREE IN NETWORK VIEWS IS NOT POSSIBLE WITH SINGLECLICK ON LARGE DOMAINS.
-----------------------------------------------------------------------------------------------------------
The base issue was in the Apache ECS library, it was failing to guarantee the ordering of items added to html elements.
The more items added the larger the potential for an item to end up being corrupted.
On the test case tree over 3500 lines of script are generated and added, this nearly always resulted in a line being over written by a later addition. A method of generating the script was improved to overcome limitations of ECS library.

IZ11700 55076 SNMP NOT COMPLETING ON CERTAIN DEVICES - RUNS ENDLESS SNMP WALKS
-----------------------------------------------------------------------------------------------------------
The fix was to update the snmp stack to perform more rigourous checking of the snmp indices returned and exit once looping was detected.

IZ12552 55160 WRONG EVENTS GENERATED DURING STARTUP OF PRECISION
-----------------------------------------------------------------------------------------------------------
The excess processing within the main Rendezvous thread has been resolved by performing it in a later thread.

IZ12734 55174 MISSING PORT CHANNEL CONNECTIONS
-----------------------------------------------------------------------------------------------------------
This Agent (CiscoRouterLayer2InfoTelnet ) is used to discover layer 2 information on Cisco router. basically This agent discover the port Channel related interface Data. it run the commands "show interfaces port-channel " and ôshow etherchannel summary" to discover port channel data.

IZ13108 55196 INTERFACE AGENT SHOULD HAVE HIGHEST PRECEDENCE.
-----------------------------------------------------------------------------------------------------------
It is possible for the m_LocalNbr->m_IpAddress and m_LocalNbr->m_Subnet fields to be set inconsistently in the workingEntities.finalEntity table.
This is because of the way ncp_disco creates the final entities by merging all the agent returns.
To avoid this, the out-of-the-box default precedence for the Interface agent has been increased from 2 to 5.
This change to the out-of-the-box default will NOT alter existing domain specific config files, so users who are worried that this problem might affect them should manually increase the precedence of the Interface agent in their domain specific config files.

IZ13912 55262 NCP_MODEL ASSUMES HSRP GROUPS ARE 'ELEMENTS' AND GENERATES "SETDISPLAYLABEL: FAILED" ERROR MESSAGES IN THE LOGS
-----------------------------------------------------------------------------------------------------------
Code changes were made to remove the warnings from the log file.

55342 Turn on compiler optimisation in AIX builds
-----------------------------------------------------------------------------------------------------------
Compiler optimisation has been enabled for 3.7 fix pack 3 on AIX.

55403 GetDiscoCache.pl doesn't tar up cache files on linux
-----------------------------------------------------------------------------------------------------------
Previously, on linux, the GetDiscoCache.pl script would faill to package a tarball from the created cache files. This is no longer an issue - a tarball will be created in the current working directory called Disco...tar.
Addiontally, the script now makes some basic checks that disco it can access ncp_disco, and attempts to automatically set the latency to an appropriate value to cope with large topologies.

IZ15451 55483 MYSQL JDBC CONNECTION POOL EXHAUSTED
-----------------------------------------------------------------------------------------------------------
Monitoring GUI was fixed in order to return all connections properly to the pool.

55487 Disco Admin GUI: Optimize translations.ipToBaseName query for Status->Discovery Devices tab.
-----------------------------------------------------------------------------------------------------------

IZ15484 55495 EXCESSIVE MEMORY USAGE IN DISCO ON AIX
-----------------------------------------------------------------------------------------------------------
A number of changes have been made to fix this issue.
1. Modification to the SendTopologyToModel process to compress the ROMP packets sent from disco to model.
2. To disable the use of our custom allocator code on AIX
3. On AIX use AIX malloc buckets instead of our custom allocator by use of the MALLOCOPTIONS environment variable.
Now AIX memory usage is the same as solaris.

55505 Topology update only triggers one user tree to update
-----------------------------------------------------------------------------------------------------------

IZ15953 55573 SNMP Helper should wait longer than the standard timeout/retries configuration when dealing with devices that have responded in the past
-----------------------------------------------------------------------------------------------------------
Implemented an exponential back-off for devices we have previously got an SNMP response from.
The number of extra retries is specified in the snmpHelper.configuration table, which can be modified in DiscoSnmpHelperSchema.cfg.
The new behaviour is basically: if a device doesn't respond within the standard timeout/retries, we begin the exponential back-off - we keep retrying the request, but double the timeout each time.
If at the end of this time, we still don't get a response, then we time out as normal, but if we get a response then we resume the operation. This behaviour only applies to table traversals by the SNMP helper. It doesn't currently affect the ncp_poller process.


Known Issues with V3.7 fix pack 2 (Build 3.7.0.19)
==================================================
52195 Disco Admin GUI: Saving discovery configuration and other GUI settings
IZ00716 52282 AMOS OVERWRITES CLEARED EVENTS WITH SEVERITY 5 WHEN LINKUP AND LINKDOWN INTERFACE EVENTS FLOWS QUICKLY
IZ03719 53360 DATA FROM NAT TEXT FILE CAN BE IGNORED WHEN OTHER NAT AGENTS ARE ENABLED.
IZ04128 53487 NCP_DH_TELNET DOESN'T ALWAYS CORRECTLY HANDLE SERVER INITIATED CONNECTION CLOSURE
IZ07978 54471 SNMP HELPER CAN MIS-FORMAT OCTET STRINGS WITH NON-PRINTABLE CHARACTERS


Summary of Fixes in V3.7 fix pack 2 (Build 3.7.0.19)
====================================================
IY89596 47379 PRECISION SEEMS TO MIX UP ROOTCAUSE/SYMPTOM FOR CERTAIN OSPF TR APS: EX: OSPFSTATECHANGE / SNMPTRAP-LINKDOWN.
-----------------------------------------------------------------------------------------------------------
The gateway configuration files NcoGateSchema.cfg/NcoGateInserts.cfg and the Nckl rules files IETF-OSPF-TRAP-MIB.include.snmptrap.rules and PreClass.snmptrap.lookup were all enhanced so that Precision (esp. AMOS) processes certain OSPF traps accurately during RCA. Note that the changes to the NcKL rules are shipped as part of NcKL v1.4.

IY99016 51444 NCP_STORE CORE DUMPS UPON FAILING SANITY VALIDATION DURING DATA MIGRATION FROM DISCOVERY DATABASE TO MODEL
-----------------------------------------------------------------------------------------------------------
A fix is provided by allowing ncp_store to ignore the same record and log where validation is failing and continue further processing instead of aborting.

52310 Error message: 'Undefined symbol ARG_1 referenced in stitcher IpToBaseName' when running Collector based discovery
-----------------------------------------------------------------------------------------------------------

52572 Need horizontal scroll bar in Network View for Network View Tree
-----------------------------------------------------------------------------------------------------------

52687 [SVT] topoviz - view create tab title switches tabs
-----------------------------------------------------------------------------------------------------------

52727 Discovery Config GUI - order of DNS options not retained
-----------------------------------------------------------------------------------------------------------

52740 Disco Admin GUI: Problem with boolean values in the filter builder
-----------------------------------------------------------------------------------------------------------
There is a problem with the filter builder in the disco config GUI and boolean values. It displays 0 as "True" and 1 as "False". For example, the "SNMP access only" filter should let through devices where m_HaveAccess <> 0, and the "Advanced" tab shows it's doing this (see advanced.jpg attached). But on the "Basic" tab it looks like it's doing the opposite (see basic.jpg attached). If you change the boolean to "False" in the dropdown then the number on the advanced tab changes to 1. To avoid this problem, check and correct filters that reference boolean fields in the "Advanced" tab before saving them.

52795 Reconnection message with database doesn't come in log files in debug 0
-----------------------------------------------------------------------------------------------------------

52873 "HTTP Status 404-" error on MIB tree page when creating/editing poll definitions
-----------------------------------------------------------------------------------------------------------

IZ02090 52911 SWITCH STITCHER CAN BE INACCURATE WHERE CDP LINKS RESOLVE TO NON-UNIQUE MAC ADDRESSES
-----------------------------------------------------------------------------------------------------------
Switch stitcher (switchFdbToConnections.stch) modified to improve the use of CDP information during connectivity resolution.

52924 [SVT] ncp_model from BACKUP PIP server got error message CMdlManagedStatus.cc lin 617 An unexpected nil pointer reference has been found
-----------------------------------------------------------------------------------------------------------

52957 [SVT] ncp_virtualdomain does NOT exit when ncp_ctrl is killed
-----------------------------------------------------------------------------------------------------------

53343 Get request for multiple objects only retrieves one object
-----------------------------------------------------------------------------------------------------------

53390 Memory leak in CRivTable.cc
-----------------------------------------------------------------------------------------------------------
Fix to minor memory leak in the database system.

53410 Poller failover needs to work with a remote backup server
-----------------------------------------------------------------------------------------------------------

IZ03912 53436 AMOS CORE DUMP DURING TOPOLOGY UPDATE.
-----------------------------------------------------------------------------------------------------------
AMOS could abort during topology update if the previous topology contained records that had both RelatedTo and UpwardConnections referring to the same entity. This is a consequence of an abnormal discovery of a device(s) Another ticket has been raised to fix the corrupt topology record issue caused by the discovery process.

53574 Change build system to work on Redhat 4 Internal - don't release note
-----------------------------------------------------------------------------------------------------------

53605 HuaweiMPLSTelnet agent produces incorrect data on devices partially lacking RT data
-----------------------------------------------------------------------------------------------------------
HuaweiMPLSTelnet agent now supports VPN configurations that don't have both import and export route targets

53640 ncp_mibinit crashes parsing 'Symbol Technologies' MIBs
-----------------------------------------------------------------------------------------------------------

IZ04934 53649 SYSUPTIME SHOWING AS NEGATIVE NUMBER IN DEVICE STRUCTURE VIEW
-----------------------------------------------------------------------------------------------------------
Scripts updatePollerDb and updatePrecisionIPDb can be found in $PRECISION_HOME/scripts/sql//, and can be used to modify some existing database tables. To run these, copy the file for the appropriate DB to the DB server, and follow the instructions given at the top of the file. The former provides performance improvements for the MySQL poller database tables, and is recommended for all users. NOTE: this file was previously found in $PRECISION_HOME/scripts. The latter corrects the types of some NCIM integer fields (sysUpTime, ifSpeed, ifHighSpeed) so that they are not incorrectly displayed as negative numbers, and is an optional modification. Note that the $NCHOME/etc/precision/ModelNcimDb..cfg must also evaluate the 3 NCIM fields as type long64 (as opposed to long or int). For example: sysUpTime = "eval(long64, '&ExtraInfo->m_SysUpTime')",

53673 Fix to NCP Webtool Template module to remove non-fatal O_RDONLY warning messages
-----------------------------------------------------------------------------------------------------------

53750 SNMP helper does not perform 'partial walk' requests properly
-----------------------------------------------------------------------------------------------------------
An error has been fixed where partial SNMP table traversals were being requested by discovery agents, but not being performed by the SNMP helper. This potentially has a significant affect on discovery performance.

53783 Disco Admin GUI: Modified Scope automatically disappeared
-----------------------------------------------------------------------------------------------------------

53796 SNMP helper not encrypting passwords during queries
-----------------------------------------------------------------------------------------------------------

53810 ncp_poller needlessly downloads managed status for entities in other domains
-----------------------------------------------------------------------------------------------------------

IZ05528 53865 ITNMIP 3.7 DESKTOP NON FUNCTIONAL USING IE6 AND JRE 1.6
-----------------------------------------------------------------------------------------------------------
A new property has been added to topoviz.properties. When topoviz.client.plugin.ieautoactivate is set to false the applet is loaded in an alternative way allowing it to load correctly in these environments.

IZ05960 53947 UNACCEPTABLY SLOW TOPOVIZ PERFORMANCE OF 3.7
-----------------------------------------------------------------------------------------------------------

54006 Collector Layer 3 topology unused, so L3 view based purely on subnet based topology data
-----------------------------------------------------------------------------------------------------------
Fixed issue where Collector based discovery lost layer 3 topology if no subnet data was present.

54031 Policies intermittently faliing to monitor entities
-----------------------------------------------------------------------------------------------------------
It was possible for newly discovered devices to be erroneously omitted from the polling schedule. The algorithm for selecting devices has been corrected so that none are missed. It relies on the domainMgr.lastUpdated field in the NCIM database being updated when changes to the topology are made.

IZ06595 54127 INSTALLER DOESN'T CORRECTLY HANDLE A DISTRIBUTED ENVIRONMENT WHERE WEBAPPS ON ONE MACHINE AND CORE COMPONENT ON ANOTHER.
-----------------------------------------------------------------------------------------------------------
Distributed installations of ITNM/Precision IP would not function as they should.
To fix the issue, once you have installed the baseline WebApplications package (on SERVER-B), immediately apply this fixpack to your installation (of the Web Applications package on SERVER-B).

IZ06760 54179 SWITCH STITCHER CAN WRONGLY HANDLE M_LOCALNBRPHYSADDR OF ' ' RESULTING IN CONNECTIVITY LOSS
-----------------------------------------------------------------------------------------------------------
Improved switch stitcher link resolution where link is resolved to an interface with a "" MAC address.

IZ06739 54182 NCP_POLLER REPORTING DEVICE DOWN DUE TO PING REPLIES HAVING CRCS
-----------------------------------------------------------------------------------------------------------
ncp_poller has been modified to add a payload and padding to a minimum ICMP packet length of 40 bytes, so that it is consistent with the old monitor system.

54222 Disco GUI: New filter window fails to load correctly on IE
-----------------------------------------------------------------------------------------------------------

54326 ncp_dh_snmp and ncp_poller processes should invalidate SNMP access details that are no longer configured to be used
-----------------------------------------------------------------------------------------------------------
If the SNMP helper has assigned access details to a device, it will continue to use those until they stop working, even if the details being used have been removed from the configuration. This fix causes the SNMP helper to reset any details at startup that it has cached, but are no longer valid, according to the configuration in snmpStack.verSecurityTable.

IZ07463 54336 CHANGING POLL POLICY DEFINITION SOMETIMES CAUSES ERRORS
-----------------------------------------------------------------------------------------------------------
The solution for 3.7 is for us to explicitly delete the rows in code that would otherwise be deleted by the triggers.

IZ07288 54894 NCP_MODEL SHOULD SYNC "MANAGEDSTATUS" VALUE FROM EACH DISCOVERY/RE-DISCOVERY CYCLE.
-----------------------------------------------------------------------------------------------------------
The TagManagedEntities stitcher will now update the managed status of existing entities if and only if it involves a transition to or from state 2. So this state - 'permanently unmanaged' - can also be thought of as 'unmanaged by disco'. Toggling between states 0 and 1 is still performed only via the GUI. When the managed status of a chassis is changed, the backend will set that status for all contained entities (as before), but if and only if the current status is not 2. In other words, entities with a managed status of 2 are now unaffected by changes to the maanged status of a containing entitiy.


IZ07287 54344 LOGICAL CONNECTIONS TO VLANS SHOULD NOT BE VISUALISED IN TOPOVIZ
-----------------------------------------------------------------------------------------------------------
Changes to the ncp_model binary were made to prevent internal VLAN connections being written into the NCIM topology.

54350 deadlock can sometimes occur in ncp_poller
-----------------------------------------------------------------------------------------------------------

IZ07917 54443 CONFIGURATION IN NCMONITOR DEFAULTPING TABLE IS NOT HONOURED
-----------------------------------------------------------------------------------------------------------
The timeout and retries is now configurable at the poll definition level for Default Chassis Ping and Default Interface Ping poll types. This configuration has been made available in a poll type specific tab in the poll definition editor for these poll types. These parameters are used for all targets in scope for a policies based on the poll definition.

IZ07799 54444 SNMP HELPER CAN STOP LONG SNMP WALKS EARLY LEADING TO INCOMPLETE DISCOVERY
-----------------------------------------------------------------------------------------------------------
The problem was caused because timeouts were being incorrectly handled in long SNMP walks. Rather than the timeout being applied per packet sent during the walk, it was often applied to the entire walk. So, if the SNMP timeout was set to 3000 milliseconds, if the entire walk took longer than 3 seconds it would often be treated as having completely failed. This bug was introduced in version 3.7 fix pack 1, and made it impossible to do an accurate discovery on any realistic network. The timeout bug is now fixed. The timeout applies per packet in an SNMP walk, not for the entire walk.

54635 ncp_poller uses 'public' community string when it is not configured to do so
-----------------------------------------------------------------------------------------------------------
The TagManagedEntities stitcher will now update the managed status of existing entities if and only if it involves a transition to or from state 2. So this state - 'permanently unmanaged' - can also be thought of as 'unmanaged by disco'. Toggling between states 0 and 1 is still performed only via the GUI. When the managed status of a chassis is changed, the backend will set that status for all contained entities (as before), but if and only if the current status is not 2. In other words, entities with a managed status of 2 are now unaffected by changes to the maanged status of a containing entitiy.

54803 SNMP helper terminates unexpectedly
-----------------------------------------------------------------------------------------------------------

IZ09356 54812 RESOLUTION OF SWITCH CONNECTIVITY ISSUES WITH END NODE CONNECTIONS AND SWITCH CONNECTIONS TO LOGICAL SWITCH INTERFACES
-----------------------------------------------------------------------------------------------------------
The fix resolves the issue by using what data is available to correctly deduce the physical switch interface or node link. The switch stitcher is still sensitive to poor data but its resilience is improved slightly.

54838 Change viewCache db table to improve performance in GUI
-----------------------------------------------------------------------------------------------------------
The GUI performance can be improved by changing the ncpgui.viewCache table to use a MyISAM database engine. This patch can be applied to a running instance of MySQL by executing the following script:
$NCHOME/platform//mysql5.0/bin/mysql ncpgui -u root < $PRECISION_HOME/scripts/sql/mysql/updatePrecisionGUIDb.sql
should be set as follows:
For Solaris : solaris2
For Linux : linux2x86
For AIX : aix5
For Windows : win32
Additionally, on Windows replace $NCHOME with %NCHOME% and $PRECISION_HOME with %PRECISION_HOME%.

54891 [SVT] initial disco appear hung agent log msg Failed to get Ip table from device with SNMP access - [ipv4 addresses]
-----------------------------------------------------------------------------------------------------------

IZ09735 54921 NON-RECURSIVE SNMPGETNEXT RESULTED IN MULTIPLE REPLIES NBS054754
-----------------------------------------------------------------------------------------------------------
The TagManagedEntities stitcher will now update the managed status of existing entities if and only if it involves a transition to or from state 2. So this state - 'permanently unmanaged' - can also be thought of as 'unmanaged by disco'. Toggling between states 0 and 1 is still performed only via the GUI. When the managed status of a chassis is changed, the backend will set that status for all contained entities (as before), but if and only if the current status is not 2. In other words, entities with a managed status of 2 are now unaffected by changes to the maanged status of a

54950 Bug in code using strncasecmp function which is unavailable in windows
-----------------------------------------------------------------------------------------------------------
The function strncasecmp is not avail in windows. Another instance of it had to be used.




Known Issues with V3.7 fix pack 1 (Build 3.7.0.17)
==================================================

IZ04128 NCP_DH_TELNET DOESN'T ALWAYS CORRECTLY HANDLE SERVER INITIATED CONNECTION CLOSURE
IZ03912 AMOS CORE DUMP DURING TOPOLOGY UPDATE.
IZ03719 DATA FROM NAT TEXT FILE CAN BE IGNORED WHEN OTHER NAT AGENTS ARE ENABLED.
IZ02090 SWITCH STITCHER CAN BE INACCURATE WHERE CDP LINKS RESOLVE TO NON-UNIQUE MAC ADDRESSES
IZ00716 AMOS OVERWRITES CLEARED EVENTS WITH SEVERITY 5 WHEN LINKUP AND LINKDOWN INTERFACE EVENTS FLOWS QUICKLY
IY99016 NCP_STORE CORE DUMPS UPON FAILING SANITY VALIDATION DURING DATA MIGRATION FROM DISCOVERY DATABASE TO MODEL


Summary of Fixes in V3.7 fix pack 1 (Build 3.7.0.17)
====================================================

IY88491 43562 DEVICES NOT BEEN FOUND/CONNECTED IN TOPOVIZ HUAWEI NE08E AND CISCO 3745
-----------------------------------------------------------------------------------------------------------
The HuaweiSwitchTelnet agent's Mediation Filter was changed to prevent discovery if the forwarding database data is not present via telnet.
This means any devices running VRP pre-v3 will be processed by the HuaweiSwitchTelnet agent, while the StandardSwitch will process post VRP 3 devices.

48583 Ncim Filter Builder has wobbly buttons
-----------------------------------------------------------------------------------------------------------
We were using a style on the buttons when they were in the normal position, however this style was not in the linked master.css.
This has now been rectified.

IY97225 50640 TIMEDSTITCHERS NEED LOG FILES W/O USING INSERT TO MONITORSCHEMA
-----------------------------------------------------------------------------------------------------------
The ncp_monitor binary now launches (via ncp_ctrl) ncp_m_timedstitcher with a -debug option. The debug level will be that of ncp_monitor. So any value assigned to ncp_monitor's -debug switch in CtrlServices.cfg will also be used by ncp_m_timedstitcher processes.

51700 The device discovered on partial rediscovery is not getting added to the dynamic view subnet.
-----------------------------------------------------------------------------------------------------------
Previously, a completely new device found on a partial rediscovery would not be added to the subnet dynamic views.
This has now been rectified.

51722 register_all_agents.bat changes directory when completed.
-----------------------------------------------------------------------------------------------------------
The register_all_agents.bat (Windows only) now does not alter the current directory.

IY99713 51780 MPLS AGENTS SHOULD CREATE IMPOSITION DATA ENTRIES EVEN FOR LABEL-LESS PATHS
-----------------------------------------------------------------------------------------------------------
The product has been modified so that the MPLS agent will not drop the MPLS paths where there is not a label.

51999 logging - stitcher rule failed to complete correctly. found in file CRivStitcherRuleFor.cc at line 263
-----------------------------------------------------------------------------------------------------------
If there is a device on the network that returns empty strings as a response to an SNMP GET NEXT on the ipAdEntAddr MIB variable instead its IP addresses, the error message printed by ncp_disco was not very informative. It has been improved.

52525 Network View should handle erroneous input
-----------------------------------------------------------------------------------------------------------
Network views did not appear when the status.tree.updateperiod property in the $NCHOME/etc/precision/status.properties file was set to an invalid value.
Topoviz has been altered so that if the value in the properties file is invalid, a default value of 20 is used.

52541 disco's instrumentation.ipAddresses not being refreshed on Full Redisco
-----------------------------------------------------------------------------------------------------------
Fix issue where on FullDiscovery, GUI discovered IPs, subnets doesn't update.

52595 SONMP agent can core dump
-----------------------------------------------------------------------------------------------------------
The SONMP agent has been changed so that it can cope with ifName's of the form "Slot x, Port y" and "x/y". Previously, it only used to work with "Slot x, Port y".

52688 ncp_disco coring consistently on RH 3.0 update 9 when LD_ASSUME_KERNEL is set
-----------------------------------------------------------------------------------------------------------
Previously, if you were running Precision on Redhat 3 update 9, you had to make sure the LD_ASSUME_KERNEL environment variable was not set before starting any of the Precision processes.
The wrapper scripts for the Precision processes now take care of this.

52744 Bug in processing of filters including defined lists i.e. IN ( ) queries.
-----------------------------------------------------------------------------------------------------------
Previously, filters of the form:
someField IN ( listItem1, listItem2, ..., listItemN )
were not processed correctly by the back-end processes. This meant, for example, that you could not have an instantiation filter:
ExtraInfo->m_IfType IN ( 6, 15, 39 )
The filter processing now handles the IN operator followed by a list correctly.

52777 LingerTime should not be decremented for unmanaged entities
-----------------------------------------------------------------------------------------------------------
Unmanaged chassis will not be removed from the topology

52781 ncp_poller core dumped
-----------------------------------------------------------------------------------------------------------
The ncp_poller process could core dump on shutdown. It no longer does this.

52807 SNMP GetNext operations sometimes hang on Windows
-----------------------------------------------------------------------------------------------------------
If, at the start of discovery a device responds to SNMP, but later stops responding, it was possible (only when running Precision on Windows) for SNMP GetNext operations to continue indefinitely (or until the device responds again).
This problem has been resolved.

52941 backup ncp_poller not fully initialized on initial startup
-----------------------------------------------------------------------------------------------------------
In Precision failover configuration, the backup ncp_poller process is now correctly initialised.

52986 Add RPR discovery capability to Precision
-----------------------------------------------------------------------------------------------------------
A new agent, StandardRPR, has been added to Precision. This can discover the connectivity of devices on resilient packet rings (RPR) that support the IEEE 802.17 RPR MIB.

IZ02412 53007 COMPAQ PROLIANT BL P-CLASS GBE INTERCONNECT SWITCH SWITCHES CAN RETURN MASSIVE AMOUNTS OF SNMP DATA AND MAKE NCP_DH_SNMP CORE
-----------------------------------------------------------------------------------------------------------
Compaq ProLiant BL p-class GbE interconnect switches are no longer analysed by the StandardSwitch agent. This is because they can make the SNMP helper go into an infinite loop.

53106 DISCO SNMP discovery only looks at the first community string in the PASSWORD table
-----------------------------------------------------------------------------------------------------------
When more than one community string was defined for a particular device, discovery was only ever trying the first community string for discovery of the device.
Now, as expected, it will try each in turn until it finds one that works.

53164 Need to include the 2nd round of translated PII packages.
-----------------------------------------------------------------------------------------------------------
Translations of text displayed in the GUIs into languages other than English have been updated.

53230 Add freeform XML support to the EMS integration mechanism
-----------------------------------------------------------------------------------------------------------
Added support for freeform XML data in Collector XML-RPC responses. See CollectorDeveloperGuide V1.2+ for details.

53243 Disco Scoping logic failing to exclude a dev specified through DiscoScope.cfg file
-----------------------------------------------------------------------------------------------------------
Individual IP addresses excluded from the discovery scope were not correctly excluded.
This problem is now resolved.

53250 memory leak in SNMP link state monitor
-----------------------------------------------------------------------------------------------------------
There was a memory leak in the SNMP link state monitor (libNcpMonitorSnmpLinkStatus.so). This has now been fixed.

53275 Fixing IPv6 issues in scoping.
-----------------------------------------------------------------------------------------------------------
Fix for scoping issues with IPv6 scope definitions.

53312 MySQL usage by ncp_dh_snmp is slowing down phase 1 discovery times
-----------------------------------------------------------------------------------------------------------
The phase 1 discovery time can be improved by removing unnecessary indexes from the poller database. To do this, run:
$NCHOME/platform//mysql5.0/bin/mysql ncpoller -u root < $PRECISION_HOME/scripts/sql/mysql/updatePollerDb.sql
should be set as follows:
For Solaris : solaris2
For Linux : linux2x86
For AIX : aix5
For Windows : win32
Additionally, on Windows replace $NCHOME with %NCHOME% and $PRECISION_HOME with %PRECISION_HOME%.
In order to confirm whether or not the script from NBS 53312 has been executed you'll need to execute the following MySQL command:
show create table ncpoller.job;
Which should show something like the following:
CREATE TABLE `job` (
`job_id` int(11) NOT NULL auto_increment,
`name` varchar(250) collate utf8_bin default NULL,
... lines cut ...
KEY `nextjob_success_idx` (`nextjob_success`)
) ENGINE=MyISAM DEFAULT CHARSET=utf8 COLLATE=utf8_bin
The bit one is interested in is the ENGINE type. If it says InnoDB then the script has not been executed.

53371 Update Precision IP Help Files
-----------------------------------------------------------------------------------------------------------
Online help available in the GUIs has been updated.