When reporting a problem to IBM Service, you might be asked for a detailed
trace for the Db2® cluster file
system,
which is based on IBM® Spectrum Scale.
Use the following information to take a IBM Spectrum Scale trace.
About this task
Enabling the trace facility (OFF by default) might impact
the performance of your system. As a result, use the trace facility
only when directed by IBM Technical
Support. The optimal values that are to be used for the trace buffer
size, file size, and tracing levels vary depending on the type of
problem. IBM Technical
Support analyzes the problem and gives recommended values for these
variables.
The mmchconfig command can be
used to configure IBM Spectrum Scale tracing
parameters.
The configuration changes are automatically applied
to all nodes in the cluster.
Procedure
To obtain a IBM Spectrum Scale trace,
run the following steps as the root user:
- Make sure the /tmp/mmfs directory
exists on all nodes.
Trace reports and internal dumps are
written to this directory.
- Set up the trace level on each IBM Spectrum Scale cluster
by entering the following command:
mmchconfig trace = trace_hooks
where
trace_hooks specifies
the type of event to trace.
Note: Values for the trace hooks depend
on the problem that is being experienced. Contact IBM Technical support for recommended
values.
- Increase the size of the trace buffer and trace file by
entering the following command:
mmchconfig traceFileSize = filesize
where
filesize specifies
the output trace log file size
Note: The value for filesize depends
on the problem that is being experienced. Contact IBM Technical support for recommended
values.
- To start the trace run the mmtrace command
on all hosts that are affected by the problem.
- To check whether the trace is started run ps -ef | grep trace.
The command /bin/trace might be displayed.
- Reproduce the problem that you want to trace.
- Dump the trace buffer by entering the mmtrace command
on each IBM Spectrum Scale node.
- Turn off the trace by entering the following command: