If more than one process is affected, try to determine which process is causing the bottleneck.

If you have already determined which process is at the center of the problem, go to the section on "Capturing Diagnostic Information."

If one process was spending most of its time in the COM state, that is the process to analyze; reduce its priority to zero (0) and go to the section on "Capturing Diagnostic Information."

For each process that is in RWAST, issue a SHOW PROC/QUOTA/OUT=file-spec/ID=pid to an appropriate file.

Perform a CCE DUMP/DB and try to stop these processes. For information on stopping processes, refer to the section of that name in this chapter. If the RWAST process(es) do not respond to all stopping procedures, perform a normal system shutdown and restart the system.

If many processes running the same application are unresponsive and in the HIB state, determine which database has the problem by running LKE SHOW /ALL. LKE SHOW /ALL reports on the regions that it has processed.

If LKE SHOW /ALL reports on all regions, then repeat the command and analyze the reports for evidence of a deadlock or a LOCK that is "stuck." Stop the problem process(es), and correct the application.

If LKE hangs, run GDE SHOW /REGION to determine the next region after the last region reported; this is the region with the problem.

Run DSE to determine what process has the critical section of the hung region. In DSE enter the following commands.

        DSE>FIND/REGION=region-name
        

then

        DSE>CRIT
        

If CRIT reports that no process owns the critical section, issue a

        DSE>BUFFER
        

command to test whether the region is OK. If BUFFER works, the database is not the problem. Restart the analysis process with the section "How many jobs are affected?" If it hangs for one minute and reports that it cannot flush, issue a

        DSE>CRIT/SEIZE
        

followed by

        DSE>CRIT/RELEASE
        

If CRIT /RELEASE hangs, go to the section on "Capturing Diagnostic Information," and use another process to analyze the process that is running DSE. If CRIT /RELEASE works, then try to

        DSE>EXIT
        

If EXIT hangs, go to the section on "Capturing Diagnostic Information," and use another process to analyze the process that is running DSE. If EXIT works, then reinvoke DSE, issue a

        DSE>CRIT/SEIZE
        

then use another process to stop the process running DSE. If this clears the problem, report the problem and return to normal operation.

If the CRIT command displays the PID of a process that has write ownership, that is the process to analyze; go to the "Capturing Diagnostic Information" section.

If the CRIT command displays a count of READ owners, run CCE DUMP /DB and go to section H5.