 |
The Question is:
Using Monitor, I've seen as many as nine processes in "CUR" state at one time.
How is this possible when the node only has six CPU's?
The Answer is :
MONITOR does a sequential scan of the process list; it does not take an
instantaneous snapshot and it does not take out a lock on the data
structures. On a very busy system, it is quite possible that the state
of previously scanned processes has changed by the time MONITOR gets to
the end of the list -- a previously scanned process has changed from CUR
to COM, and a yet-to-be scanned process has changed from COM to CUR by
the time MONITOR gets to it, for instance. MONITOR displays the result
of the entire scan when it has completed the pass. The process state
displayed by MONITOR is thus the individual state of each process from
the recent past, and is not a snapshot of the aggregate state of all
processes.
Also do not forget that MONITOR itself is running in a process, and is
subject to being pre-empted by other processes. So the MONITOR process
scan can take a relatively long time, which also contributes to the
behavior you are seeing. You can also see odd data resulting from
processes performing periodic work and MONITOR -- values for activity
can be unusual, depending on the particulars of the process activity
and the MONITOR sampling interval.
|
|
|
|
|
