Known trace problems and limitations

Edit online

The known problems and limitations of trace processes are gathered to help you resolve problems that you might encounter when you are running a trace process.

If tracing is not currently active for a process and dsmtrace is used only with the -TRACEFLAGS option, for example, dsmtrace enable 2346 -traceflags=service, then you still see the following message:
```
ANS2805I Tracing has been enabled.
```
In this case, the trace flags were enabled, but tracing is not active until a trace file is specified by using the -TRACEFILE option.
Do not use the dsmtrace enable command to start tracing the application programming interface (API) for Data Protection applications if the Data Protection application is run in a manner that does not cause it to connect to the IBM Storage Protect server. For example, the Data Protection for HCL Domino command line interface has several such commands:
- domdsmc help
- domdsmc set
- domdsmc query domino
- domdsmc query pendingdbs
- domdsmc query preferences
If you use dsmtrace to enable tracing for such commands, the result can be a stoppage of the dsmtrace process and (AIX® and Linux® only) a residual named pipe in the /tmp directory.
You must be logged in with a local administrative account to use dsmtrace.
You must be logged in as root to use dsmtrace. If a client process stops or is stopped, it might leave a named pipe (UNIX FIFO) in the /tmp directory. These FIFOs have names that begin with TsmTrace and they include a process ID (PID) number. If a client process stops or is stopped, and then a new client process is started whose PID happens to match that of the old residual FIFO, then the trace listener thread might not start. Any old FIFOs with process numbers that do not match the FIFOs of running the IBM Storage Protect processes can be safely deleted. Do NOT delete the FIFO of a running process.
The threading model for some versions of Linux is to run each thread as a separate process, meaning that when you query process information, you might see several processes for each instance of the client. The process that you need to identify is the dsmc parent process.
When multiple instances of the same program are running, you must identify the PID of the instance that you want to trace. In such a situation, information such as process information from the operating system might be available to help you identify the required PID. For example, if you want to trace dsmc that is being run by user 'andy' and there are two instances of dsmc, one owned by user 'andy' and the other owned by user 'kevin', you can use the process owner to identify which process to trace.
If an options file contains a false option and the client does not start, you might see some named pipe errors in the dsmerror.log file. These error messages can be safely ignored.