SCF Reference Manual for J-Series and H-Series RVUs

SCF Commands
SCF Reference Manual for J-Series and H-Series RVUs—529649-003
5-126
TRACE Command
If TO file-spec is not specified, the existing trace is either stopped (if STOP is
specified) or modified as specified by trace-option .
LOCKSIZE pages
designates how many pages of memory are locked down at one time. The pages
variable is an integer in the range 4 through 1024 bytes; the default is 64. Its value
must be less than the value of PAGES pages (described below).
You can specify this option in subsystems whose IOPs perform interrupt handling,
in order to improve system performance during a trace with reduced risk of
incurring a page fault. Be aware, however, that the more memory you lock, the
greater the negative impact on system performance.
NOCOLL
optionally specifies that trace data from the extended segment is written to the file
specified in TO file-spec only when one of these situations occurs:
•
The trace is stopped.
•
The number of trace records written to the extended data segment is equal to
the COUNT specified in the TRACE command that started the trace. When
the trace facility detects this condition, the trace records from the extended
data segment are written to the file and the trace is automatically stopped. As
a result, a separate TRACE command does not need to be issued to stop the
trace.
If NOCOLL is not specified, a trace collector process reads the trace records from
the extended data segment and writes them to the file as they become available.
You cannot use NOCOLL with BULKIO.
PAGES pages
designates how much memory space, in units of 2048-byte pages, is allocated in
the extended data segment used for tracing. The
pages variable is either 0 or an
integer in the range 4 through 1024. If PAGES is omitted or if
pages equals 0, the
default value of 64 pages is assumed.
BACKUP
when given as a trace option, specifies that the backup process should receive the
trace request. The object must be running as a process pair if this syntax is used.
If the primary process is being traced when a takeover by the backup process
occurs, the trace continues. However, in this case most events being traced
before the process switch are no longer traced, because the process being traced
is no longer the primary process. If this trace option is omitted, the primary
process is assumed.