Collecting Data for Spectrum Protect: Backup/Archive client hang on Linux

Answer

Collecting data early, even before opening the Case, helps IBM® Support quickly determine if:
Symptoms match known problems (rediscovery).
There is a non-defect problem that can be identified and resolved.
There is a defect that identifies a workaround to reduce severity.
Locating root cause can speed development of a code fix.

Collecting Data: Table of contents:

	Gathering general information
	Gathering Spectrum Protect client hang Information
	Submitting Information to IBM Support
	Online Self-Help Resources
	Related Information

Manually Gathering General Information

Gather the following files:

• dsm.sys/dsm.opt
• details of operating system levels
• dsminfo.txt - Run QUERY SYSTEMINFO using the Spectrum Protect Client Command Line to generate a current version of the file.

Note: The above Spectrum Protect Client files can be found in the installation file path. The default installation path on Linux is /opt/tivoli/tsm/client/ba/bin.

Gathering Spectrum Protect Client Hang Information

From a unix shell prompt, run the following commands :

• cd /opt/tivoli/tsm/client/ba/bin
• ps -ef | grep dsm > ps.out
• pstack [pid] > pstack.out

[pid]: The process id of the hung Spectrum Protect client process.
• kill -11 [pid]

[pid]: The process id of the hung Spectrum Protect client process. This will force a core dump of the hung Spectrum Protect client process.
• ldd <proces_name>

<process_name> : name of the hung Spectrum Protect client process (i.e: dsmc, dsmcad, dsmj, ... etc)
The ldd command will list the names of the shared libraries used by the hung Spectrum Protect process.
tar and compress the listed libraries as libs.tar.Z

If the gdb tool is installed on the system with the hung Spectrum Protect process, run the following command :

• getcoreinfo <program_name> <core_file_name>
  <program_name> : name of the hung Spectrum Protect client process (i.e: dsmc, dsmcad, dsmj,...etc)
  <core_file_name> : core_file_name: core file produced by the "kill -11" command
  getcoreinfo : script attached in this document

*Notes*:
- The "kill -11" command will only produce a core file if the "core file size" ulimit is not zero.
- The "ulimit -a" command may be used to check the current "core file size" setting, A full core will not be created if the current "core " limit is smaller than the core size.
- The "ulimit -c unlimited" command may be used to set the "core file size" to unlimted.
- The getcoreinfo script must be given execute permission with the "chmod +x getcoreinfo" command.
- The getcoreinfo script generates the getcoreinfo.txt and getcoreinfo-shlibs.tar.gz files in the current directory.

If the client hang can be recreated, add the following trace options to the dsm.opt file of the client :

tracefile /myfs/tsm.trc ...modify the trace file name as appropriate for your environment.
traceflags service
tracemax 200

Recreate the hang and collect the Spectrum Protect client trace.

Provide the following files:

• ps.out
• pstack.out
• core file
• libs.tar.Z files from ldd command
• getcoreinfo.txt
• getcoreinfo-shlibs.tar.gz
• trace file (if hang can be recreated)

Submitting Information to IBM Support

After a Case is open, you can submit diagnostic troubleshooting data to IBM.

When data is uploaded to the Case, the status will be automatically be updated to acknowledge that data has been received. If you are still working with older Problem Management Records, you will want to update the PMR to indicate that you have sent the data.

Online Self-Help Resources

• Review up-to-date product information at the Spectrum Protect Support page.
• The Tivoli Storage Manager Product Support page is also still available if you have the older product.
• Sign into your account on the Spectrum Protect Product page, to access additional resources or to open a Case with the Support team.

Related Information

• Exchanging information with IBM Technical Support
• IBM Spectrum Protect Downloads - Latest Fix Packs and Interim Fixes