TNPM TIPS\TNPM 1.3.2 Known Issues

Known Issues with Tivoli Netcool Performance Manager 1.3.2

Abstract

Known issues with Tivoli Netcool Performance Manager 1.3.2, including installation and upgrade issues. This technical note outlines issues for both the Wireline and Wireless Tivoli Netcool Performance Manager components, with an added section covering all known Model Maker issues.

Content

This technical note outlines issues for both the Wireline and Wireless Tivoli Netcool Performance Manager components, with an added section covering all known Model Maker issues.

Wireline Specific Install Issues

Selection of tabs in Tivoli Integrated Portal

In the Tivoli Integrated Portal, when more than tab is open, clicking on the empty space on the tab, does not bring up the Tab. You must click on the name in the Tab to open it.

HAM installation - FTPERR appears on fresh install of High Availability Manager

If when checking the proviso.log file you see errors similar to the following:

V1:4159 2012.03.19-09.46.22 UTC FTE.1.1-10594:14390 FTPERR GYMDC10123W Unable to change remote working dir: ...//opt/dataload/SNMP.1.1/output V1:4164 2012.03.19-09.46.24 UTC FTE.1.1-10594:14390 FTPERR GYMDC10123W Unable to change remote working dir: ...//opt/dataload/SNMP.1.1/output V1:2973 2012.03.19-09.46.38 UTC FTE.2.2-10598:12917 FTPERR GYMDC10123W Unable to change remote working dir: ...//opt/dataload/SNMP.2.2/output V1:2977 2012.03.19-09.46.40 UTC FTE.2.2-10598:12917 FTPERR GYMDC10123W Unable to change remote working dir: ...//opt/dataload/SNMP.2.2/output

These errors can be ignored as they will disappear once you test HAM failover.

Testing of the High Availability Manager is discussed in "TNPM Wireline High Availability Installation and Configuration" guide, which is available on ISML:

[1]

Tivoli Common Reporting installation fails and does not notify the deployer

This is a problem with Install Anywhere. Flexera Problem Report LWIS7XNMUS: Silent install fails with zero return code when rule on installer does not pass.

Currently no release of Install Anywhere has been targeted to fix this.

For Netcool/Proviso, this means that if a Tivoli Common Reporting installation fails, it will not notify the deployer. The "Install TIP" step will report success even though there may have been errors during the installation process.

Symptoms include:

· The "Install DataView" Step fails. · The Tivoli Integrated Portal URL comes up but Tivoli Common Reporting is not available.

Suggested action:

Always check that the Tivoli Integrated Portal URL comes up with all installed components before clicking "Done" on the Deployer window.

TIP Install Step failed but Install log says "Installation Complete"Add

In the deployer, the remote TIP Install Step failed but Install log says "Installation Complete"​ ​ To view the install log, right click on the install step, select Properties and then 'Output'. The message inside says "Installation Complete"​ ​ The remote install was a failure because the tar file that was internally copied over from the main server to the remote server in the previous step "Prepare Remote TIP install" was incomplete.​ ​ This was due to the disk space issue on the main server.​ ​ Suggested action:​ ​ Solution: verify disk space before proceeding with install ​

Uninstalling DataMart on a host with a fully qualified domain name

Additional steps are required to uninstall a DataMart from a host that has a fully qualified domain name.

In this procedure, you substitute the fully qualified hostname for the short form of the hostname <hostname_short>.

I. Run the deployer for the uninstallation. II. In the wizard, compete each step until the final page containing the Step list. III. Open a terminal window on the host where the DataMart you are uninstalling is located. IV. Change to the /tmp/ProvisoConsumer/Plan/MachinePlan_<hostname_short>/*_DataMart_step/cfg directory. V. Edit the DataMart.cfg file. VI. To hardcode your host name, using the short version, in place of the place holder $TARGET_HOST, change the line TARGET_HOST="$TARGET_HOST" to TARGET_HOST="<hostname_short>". VII. Save the DataMart.cfg file. VIII. Return to the deployer. IX. Run the Uninstall DataMart step.

On Linux, if the Check DB step in the Deployer fails, ensure that the kernel parameters are correctly set

On Linux, if the Check DB step fails, ensure the kernel parameters are correctly set. Kernel parameters must be set to the recommended values for the Oracle installation. If any of the kernel parameters are incorrect, the step fails and an error message is logged with the expected value.

You can ignore the fact that the error message refers to "The Solaris semaphore" value. This is an open issue. The message should not just refer to Solaris.

You can set the correct kernel parameter by using the following command:

sysctl <param>=<value>

For example,

sysctl kernel.shmmax=27459232 (or whatever the recommended value is)

If the server has previously been used to install Proviso then the existing kernel parameters are sufficient.

RSH is not available as a transfer protocol

The Deployer allows user to select "RSH", "SSH", or "RSH and SSH" for remote execution of commands. The RSH option is not supported. You can choose either the "SSH", option or the default "SSH and RSH" option. SSH needs to be configured for both of these options.

ProvisoPackInstaller only gives space to enter one TIP server at a time

You need to use the Topology Editor to configure multiple servers to get the check list with multiple servers. If there is no topology documented configured in the database, the screen will only allow you to enter the login information for one server at a time.

You can either deploy the Technology Pack repeatedly by giving a different TIP server each time using Packinstaller or you can use the DataView’s export.sh and import.sh scripts to move the content from one TIP server to another. For more information on the export.sh and import.sh scripts, refer to the DataView User and Administrator Guide.

HAM components failed to uninstall

The uninstall process marks all steps as "Success". However, when the topology is loaded from the database, the HAM components are still present. Running the uninstall process for the second time removes only a few components.

Workaround: I. Save the current topology in the Topology Editor to a local file. For example, /tmp/HAMuninstall.xml. II. Load the local topology file /tmp/HAMuninstall.xml to the Topology Editor. III. Remove the remaining HAM components. IV. Save and deploy the topology file. V. Load topology from the database. All the HAM components should be successfully removed.

Proviso InfoD startup scripts error

When installing DataMart, the TNPM installer creates two files, one in /etc with the name pisrc3113ProvisoInfoServer.sh, and the other in /etc/rc.d/init.d/ with the name pis3113ProvisoInfoServer.

When running the pis3113ProvisoInfoServer command located in the /etc/rc.d/init.d directory, the following error is shown:

$ /etc/init.d/pis3113ProvisoInfoServer bash: /etc/init.d/pis3113ProvisoInfoServer: /usr/bin/sh: bad interpreter: No such file or directory

Even if you change the directory from #!/usr/bin/sh to #!/bin/sh in the pis3113ProvisoInfoServer script and run the script again, the following error is shown:

$ /etc/init.d/pis3113ProvisoInfoServer /etc/init.d/pis3113ProvisoInfoServer: line 95: .: filename argument required .: usage: . filename [arguments]

Workaround: The Proviso InfoD process needs to be manually started on the Linux DataMart system that is rebooted. To accomplish this, perform these steps: 1. Login to the DataMart server as 'pvuser'. 2. Source the dataMart.env file. 3. Execute the command: pismgr start

Install steps do not appear if JRE step is unchecked for a DISC on remote host

When installing a DataMart with a Discovery Server to a remote host and this is the only host being deployed, if the JRE host is unchecked in the Host Summary page, the install DataChannel steps do not appear in the Deployer. This only occurs in a High Availability environment.

Workaround: Check the JRE host in the Deployer host selection pane. If the Deployer deploys to the DataMart host and the JRE host, all steps appear.

No trailing white space in the value for DiscoHostName for a High Availability install

The value for the DiscoHostName resource in the PVMHOME/res/C/inventory.User.res file must not have any trailing white space. When there is a trailing white space in the value for DiscoHostName, the GUI fails to utilize the value properly.

Rerun the deployer after completely uninstalling the DataChannel component

When you completely uninstall the DataChannel component from your environment, the Load Channel Configuration step fails in the deployer. When the deployer completes, rerun the deployer. You are presented with the Load Channel Configuration and Load Collector Configuration steps only. Run the steps to successfully complete the uninstallation.

Unable to remove the database from the topology

If you are unable to remove the database from the Topology, run the deployer for uninstallation. You are presented with the Load Channel Configuration and Load Collector Configuration steps only. Run the steps and reload the topology from the database. The Remove action is now enabled on the Database Configurations object in the topology.

Wireline specific upgrade issues

Load_SQL_Module fails as client is still connected to the system during an upgrade of Oracle Enterprise for a Tivoli Netcool Performance Manager

When Oracle Enterprise Management GUI client is still connected to the DB, additional steps are required to upgrade Oracle Enterprise for a Tivoli Netcool Performance Manager upgrade. You must run the following steps before doing an upgrade of Oracle Enterprise for the Tivoli Netcool Performance Manager to make sure that there are no more clients connected to the DB before applying an Oracle patch:

1. Connect to the system as a Tivoli Netcool Performance Manager database administrator (default: pv_admin/pv@PV) using the sqlplus command: sqlplus pv_admin/pv@PV

2. Check for connected user sessions: Select * From V$Session Where Nvl( Username, User ) != User;

3. Close any sessions that are shown by the above query by closing the application listed. For example, close the Datachannel components and close any sqlplus sessions.

4. Check for connected user sessions again. If there are still sessions active in v$session and you cannot find the application to close, close each session with: alter system kill session 'SID, serial#'; Where SID and serial# correspond with the session listed in v$session. For example: alter system kill session (7,258);

5. Ensure that there are no connected user sessions before starting the patch process.

IP SLA Monitoring for Cisco IP SLA Technology Pack

When the collector is installed and started by the installer, some key tables are missing from the database. These tables contain configuration data that enables IP SLA monitoring to work on the Cisco IP SLA Technology Pack. After you complete the installation, if IP SLA operations are required, bounce the collectors once to reload the configuration.

You also need to manually install and compile the CISCO-RTTMON-IP-EXT-MIB.my MIB. The Management GUI will fail all refresh, list, and set commands if this MIB is not installed. You must also perform this MIB installation if you are upgrading your Tivoli Netcool Performance Manager from pre 1.3.1 to Tivoli Netcool Performance Manager 1.3.1.

If you receive a warning message about unresolved dependencies during the CISCO-RTTMON-IP-EXT-MIB.my MIB compilation, you must import and compile the following MIBs together:

CISCO-RTTMON-IP-EXT-MIB.my IPV6-FLOW-LABEL-MIB.txt INET-ADDRESS-MIB.txt DIFFSERV-DSCP-TC.txt Name the compiled MIB file as CISCO-RTTMON-IP-EXT-MIB.my.oid and place it in the /opt/dataload/mibs/oid/ directory.

After the DataView component is installed and started for the first time in Tivoli Netcool Performance Manager 1.3.1, you must stop and start all of the SNMP DataLoad component so that they can pick up the the new probe type definition from the database. Failure to find the new probe type definitions, will cause errors in the IP SLA probe management pages. Probe creations will be systematically rejected with 'invalid type' errors.

General Wireline Issues

DataMart documentation has been merged

DataMart Online help documents may still refer to the Datamart Configuration Guide and DataMart Operation Guide; however, these guides do not exist as individual guides anymore, but form one guide called the DataMart Configuration and Operation Guide.

PD4ML issue: Differences between pdfs created using the pdf icon and those sent as a scheduled report

The pdf documents created using the pdf icon and the pdf sent by mail as a scheduled report differ.

The report created using the pdf icon has resource names and values underlined and in blue. It also has the non-functional filer box below. These are absent in the reports sent via the scheduler.

PD4ML issue: Non-functional Export CSV link.

The pdf files created from any report in the defined resource view contain a non-functional Export CSV link.

The Version Registration table contains DataLoad SNMP 0.0 entries for the previous version

A Version Registration table may contain DataLoad SNMP 0.0 collector entries for a previous version of Tivoli Netcool Performance Manager. These table entries, if present, are put there by the Topology Editor during the Upgrade Process.

When the topology editor loads the topology from a previous version it creates version registration entries for the previous version for all of the secondary DL instances involved in HAM 1+1 instances. This version registration entry does not impact anything in the system, except the topology editor and deployer.

Any DL SNMP 0.0 for the previous Tivoli Netcool Performance Manager version within the version registration table can safely be ignored.

SNMP does not check for alternative Ipv4-Ipv6 addresses on DataLoad

The format “::10.127.77.7” is a deprecated IPv4 mapped IPv6 address. It is a valid address, but how the format is interpreted is up to the underlying operating system stack. The new format for IPv4 mapped addresses is “::FFFF:10.127.77.7”.

Linux and Solaris treat the “::10.127.77.7” address as a true IPv6 address. Collections do not work on these platforms unless a system has the address.

AIX treats the “::10.127.77.7” format the same as the “::FFFF:10.127.77.7” format. This allows collections to work because the IPv4 address is being used.

If you run “dialogTest2 IPCacheDump”, you can see that “::10.127.77.7” is mapped to “::FFFF:10.127.77.7” on AIX, and remains as “::10.127.77.7” on Solaris/AIX.

It is recommended that the standard IPv4 to IPv6 mapping are used on all three platforms for consistency. The format is “::FFFF:10.127.77.7”.

This issue is not a problem with Tivoli Netcool Performance Manager. This issue is caused by a difference in how different operating systems interpret certain IP addresses.

Limitation of qping

qping only supports up to 64 IP addresses on the command line at one time.

For IPv4 addresses, if there are 100 devices to the list of 'Specific elements' in the 'IPaddress include area' of the Inventory discovery tool wizard, then only 64 of these devices are discovered.

For IPv6 addresses, if there are 100 devices to the list of 'Specific elements' in the 'IPaddress include area' of the Inventory discovery tool wizard, then all 100 devices are discovered. There is no ping or SNMP message issued from the collector for the 36 undiscovered devices.

SNMP 'watchd' process cores when stopping

When the SNMP collector is stopped by the pvmdmgr stop command, the watchd process is expected to stop within a minute. The watchdog erroneously stops by throwing an exception. A core file is placed in the state directory of the dataload root. When a collector is started, if a core file is found in the state directory, a message is printed, a post mortem file is created in the log directory, and the core is moved into core.bak. It is safe in this case to ignore core.bak and related log reports from the watchd process.

Correct notation for the Rule Editor filter

The following IPv6 style addresses should not be used: %(Element.ipAddress)='0:0:0:0:0:FFFF:10.127.77.46' %(Element.ipAddress)='::FFFF:10.127.77.46' %(Element.ipAddress)='2002:092a:0111:0200:0010:0127:0077:0046'

The correct notation is as follows: %(Element.ipAddress)='2002:92a:111:200:10:127:77:46'

The vertical axis does not show up for a straight line plot in Performance Diagnostic View.

When there is a straight line plot, for example, 18 plot points with the same value 100, the vertical axis does not show up in the chart.

Performance Diagnostic View chart values are supported up to 3 decimal places

In the DataView component, values in Performance Diagnostic View charts are supported up to 3 decimal places.

Tooltips limitation in Internet Explorer

Internet Explorer throws a warning when a script is taking too long to run and complete when a chart in Diagnostic View is triggered to return 5 metrics over a 24-hour period. Similar views work with no script time out warnings in the Firefox browser.

The script timeout issue in Internet Explorer results in partial data returned and graphed. To alleviate this performance issue with Internet Explorer, the tooltips feature has been disabled for Internet Explorer. The tooltips are still active for the Firefox browser.

Follow up with Aaron on this later in the day. He is considering rolling in Paul Barry’s solution.

The TARGET HOST field is not refreshed when you select or toggle the "Change Target Host for BLB and IF to x host" option in Alcatel 5620 Bulk Load Balancer (BLB)

In the Topology Editor’s Logical View, if a user selects a BLB and uses the right-click operation to move the IV and BLB from channel to collector host or vice versa, the BLB TARGET HOST property will not automatically be refreshed once the operation is complete.

You need to click on the property or click away and then back to the BLB object again to see the refresh value. As a precaution, save the topology prior to the refresh to ensure the value in the memory is always the latest.

Cancelling topology load can leave problems view populated

If you cancel a topology load, the partially loaded topology is cleared from the Topology Editor views. However, the problem view sometimes retains a list of problems with the partially loaded topology. This issue has no functional impact.

Ignore MOVE_FAILED message for bufferedContainers file

The bufferedContainers file exists in the directory but CME is looking at elsewhere for the file.

An example of the MOVE_FAILED message in the proviso.log:

V1:12750 2010.11.25-02.02.50 UTC CME.1.2000-3867:5492 MOVE_FAILED GYMDC10421W Failed to move 'file:///opt/datachannel/CME.1.2000/SLE.2000/state/temp/state/bufferedContainers' to 'file:///opt/datachannel/CME.1.2000/SLE.2000/state/bufferedContainers', details: 'No such file or directory ("/opt/datachannel/CME.1.2000/SLE.2000/state/temp/state/bufferedContainers")'

When there is a MOVE_FAILED message in the proviso.log for the bufferedContainers file, then it can be safely ignored.

The Tivoli Integrated Portal 'iscadmins' and 'administrator' roles and the 'tnpmAdministrator' role are required for the User Preference Management page

A user must be assigned both the Tivoli Integrated Portal 'iscadmins' and 'administrator' roles and the Tivoli Netcool Performance Manager 'tnpmAdministrator' role to have the appropriate permissions to manage users by using the User Preference Management page in Tivoli Netcool Performance Manager.

For more information on the User Preference Management, refer to the DataView User and Administrator Guide.

New logging process for DataChannel requires extra step when changing log settings

In the TE, the "Log Server" the properties are available.

The property FILTER is used for the proviso.log and the property LOG_FILTER is used for the tnpm.log.

When the values are changed to "FEWI 1234" for FILTER or LOG_FILTER, restarting the dccmd and findvisual will not effect the changes as LOG is not under the findvisual process. Logging is now a JAVA process.

A user must find the java process for the LOG and kill it. Using ps -ef|grep analytics.Main will find the process to be killed. Kill this process and restart the log by running logw (the watchdog script for LOG) or wait 5 minutes for the cronjob to restart the process.

The java process will be similar to this: pvuser 26488 25744 0 18:01:59 pts/75 1:05 java -Xms256m -Xmx384m com.ibm.tivoli.analytics.Main -a LOG

HAM - when restarting pvmd, core log is created

If you kill pvmd and don't kill the watchdog, you will get a "core" dump when the watchdog stops itself.

User action: you can ignore the core dump if you can see on startup that the core dump is the result of the WatchDog process.

Oracle Client

Product : TNPM Wireline 1.3.2

Platform : Linux

Activity performd : Installation

Actual Problem faced : In installation document for TNPM Wireline 1.3.2 ; please refer below comments :

Document mentions as below :

Instructions on how to install the Oracle 11.2.0.2 client (32-bit). The Oracle client should be installed to all servers hosting Tivoli Netcool Performance Manager components, except for the server hosting the Tivoli Netcool Performance Manager database.

But in actual ,

The deployer and topology editor use the 32 bit Oracle client and the deployer *must* run on the DB server host, this forces every host in the TNPM installation to require the 32 bit oracle client

64 bit server into /opt/oracle/11.2.0 32 bit client into /opt/oracle/11.2.0-client Install the client first DM,DC, DL all use 32 bit clients.

1.3.2 might not install with 11.2.0.2 as its looking fot 11.2.0.1

Datachannel watchdogs

You can view the java LOG process with the following command :

'ps -ef | grep java' and look for the following type :

pvuser 972 971 0 Jul 22 ? 22:29 java -Xms256m -Xmx384m com.ibm.tivoli.analytics.Main -a LOG

When restarting DC components and/or launching logw daemon (or crontab launches the process), it will show up under 'ps -ef | grep log' for a brief period of time and looks like this :

pvuser 1916 1915 99 17:41 ? 00:00:01 LOGW_visual -nologo -noherald /opt/datachannel/bin/dc.im -headless -a LOGW...

RHEL Versions

Configuration recommendation guide for TNPM 1.3.1 that available at link : http://publib.boulder.ibm.com/infocenter/tivihelp/v8r1/topic/com.ibm.netcool_pm.doc/Guides/configuration_recommendations_guide/TNPM_1.3.1_Configuration_Recommendations_Guide_Wireline_Component.pdf On Page: 15 providing information that support version is RH Linux 5.5 . Could I install and use TNPM 1.3.1 with more resent version of RH Linux OS

The TNPM 1.3.1 tested and certified with RH Linux 5.5 but we don't have any restriction or limitation for installing TNPM 1.3.1 on latest available RH Linux releases such as 5.8 or 6.2