To troubleshoot an installation problem, identify your problem in the following list, and follow the link to a description of the solution.
If you cannot find your problem described in the list, and cannot resolve it yourself, contact Customer Support Services as described in Information You Should Have and Customer Support.
Problems:
For details, see Problem: The Server Starts in Safe Mode .
For details, see Problem: Java Listener Fails to Start With JVM not found Messages Written to the Log.
This means that the underlying glibc libraries are not high enough for isetup (nor the server) to run. Examine the error message to determine which version is missing, and then install that GLIBC RPM version (or higher), and any GLIBC dependencies, before proceeding.
For details, see Problem: Setting ulimit to Allow Core Dumps.
Many UNIX environments intentionally limit the ability to read core files in setuid applications (the mode in which Security Provider OPSYS runs) for security reasons. On some platforms, the feature is configurable, but the commands to activate it vary by platform or may not be implemented in earlier OS releases. If the crash can be reproduced in an unsecured server, the core information will be produced. This is the best route to producing a complete save diagnostic. If the server cannot be run unsecured to produce the crash, see the following for details:
Problem: Forcing Core Dump Information on Solaris.
Problem: Forcing Core Dump Information on HP-UX.
Problem: The server starts in safe mode. The Web Console home page displays a message stating that the server is in safe mode and describing what triggered it.
Cause: A common cause for the server starting in safe mode is a problem with the server administrator ID password. For example, the password may have been updated on the operating system but not on the server, so the encrypted copy of the password stored by the server is out of synchronization with the password on the operating system.
Solution: The server administrator can click the fix hyperlink, which is displayed under the problem description, to display the relevant pane and resolve the problem.
For example, if the problem is that the server administrator password is out of synchronization:
The Access Control pane is displayed on the right.
The Security Mode pane opens on the right.
Problem: The listener start request fails with JVM not found messages written to the edaprint.log file.
Cause: If the server cannot find the Java Virtual Machine (JVM), the JSCOM Listener will not be able to start, and messages will be written to the server log stating that the JVM cannot be found.
The log may be found in the following location:
On Windows:
drive:\ibi\srv77\product_type\edaprint.log
On UNIX:
ibi/srv77/product_type/edaprint.log
On IBM i:
/home/iadmin/ibi/srv77/product_type/edaprint.log
where:
Is the hard drive on which the directory resides (on Windows).
Designates the type of product. The default values are:
FFS | for a Full-Function Server |
DM | for a DataMigrator Server |
WFS | for a WebFOCUS Reporting Server |
WFM | for a Shared Application Server for WebFOCUS Maintain |
Solution: Set up the JVM as described in JVM Requirements for Java Services (Server Installations Only).
The ulimit value of a process controls how large (in blocks) a core can grow. If the value is set to zero, no dump is produced, and the dump information is not read, therefore, a proper save diagnostic stack trace (snap) cannot be produced.
To check the current value, issue:
ulimit -c
To set a ulimit so that dump information can be produced, stop the server, set a value, and restart:
bin/edastart -stop ulimit -c 99999 bin/edastart -start
The actual size value is in blocks and will vary by need. Since the need is unpredictable, select a number and then check the dump information. If the information is incomplete, increase the value.
Solaris uses the coreadm command to control the ability to produce core files.
To see the current value, issue:
coreadm
For secured servers, before the server starts, issue:
coreadm -e proc-setid
No reboot of OS or service daemons is required, but core files must have a non-zero ulimit –c value.
As of HP-UX 11.31, the coreadm command controls the ability to produce core files. Sites with an operating system prior to HP-UX 11.31 can only use the unsecure server method to produce a complete save diagnostic for a crash.
To see the current settings or values for machines that support the coreadm command, issue:
coreadm
For secured server purposes (on HP-UX 11.31 or higher), before the server starts, issue:
coreadm -e proc-setid
No reboot of OS or service daemons is required, but core files must have a non-zero ulimit –c value.
AIX uses the chdev command to control the ability to produce core files. This command is on by default, so it only needs to be adjusted if it has been turned off.
To see the current value, issue:
lsattr -El sys0 -a fullcore
For secured server purposes, before the server starts, issue:
chdev -l sys0 -a fullcore=true
While Linux has options to activate core dumps, none currently work in the context of the server. Linux sites can only use the unsecure server method to produce a complete save diagnostic for a crash.
When a process crashes, the operating system generally produces a core dump. The server software is designed to detect this event and use the system debugger to extract the state of the crashed process from the core and produce what is known as a snap. While the specific debugger command may vary by vendor, the standard debugger for the vendor must be installed or a Not Found condition on running the debugger can occur (effectively a core, but no snap information). Some vendors install the debugger in /bin or /usr/bin, which are normally on $PATH, but some vendors use locations not normally on $PATH. This can result in a secondary reason for a debugger Not Found condition. Once the debugger is installed and/or on the $PATH, reproducing the crash condition will then produce the snap information.
The full message indicates the need to review edapth traces for r1shmop* entries with errors. If the server was not started with traces, start it with traces, and then view the edapth trace.
One of the r1shmop* entries in the edapth trace will show a specific error, but a common error is size is greater than system shared memory limit. This particular message indicates that the system kernel value for shared memory needs to be increased. The actual required value is generally a multiple of machine page size (typically 4K, but it can vary). The number of agents a server runs, and other installed software can also be a factor, and the required value may vary (slightly) from release to release.
There are tools, such as size and ps that will allow an experienced administrator to narrow down the precise shared memory size requirements, considering all of the software in use. However, a good rule of thumb is to increase memory in 10% increments until a working value is found.
Error messages other than shared memory size can occur, in which case, the system message is displayed. These other messages may provide an administrator with enough information to determine the appropriate action. If not, call Customer Support Services for a review. Actual kernel change commands or steps vary by vendor, so they are not explicitly outlined here.
In core dump (crash) situations, the stack information may or may not provide enough information for a problem to be resolved. Debuggable versions of the software will generally provide that information, but would not normally be installed nor used due to the disk overhead they take and they are not optimized for performance.
If a diagnostic is determined to not have enough information and use of debuggables is warranted, Customer Support Services will inform you to install and activate the debuggable version of the server and re-run the reproduction to capture a new diagnostic with the detailed stack to help troubleshoot the problem.
Caution: Do not activate the debuggable version unless explicitly requested to by Customer Support Services.
To activate the debuggable version of the server:
edastart -stop edastart -dbgon edastart -start (run until repro is completed) edastart -stop edastart -dbgoff edastart -start
Customer Support Services will provide you with additional instructions as your situation requires.
iWay Software |