Monitor FAQ for version 2.0
Last updated:  Feb 2001

Q: Disk statistics do not show any busy percentages anymore.
A: The magical answer is to run sar for a couple of hours. Somehow
   SAR activates 'disk active time' counters. 
   You probably have to do that after every reboot.

Q: Every our the disk stats in the logfile show strange out of
   range values (negative values).
A: You probably have SAR running constantly. It seems SAR resets
   (disk) counters every hour, or every time it starts!

Q: Are there any issues with monitor 2.1.2 and the year 2000?  (i.e. is 
   it year 2000 Y2K compliant)? 
A: All monitor version are y2k complient as long as the underlying OS and
   hardware is.
   Note: We cannot be held responsible for any negative results from using 
         monitor, even if it is related to Y2K!

Q: The monitor displays some big numbers like 
   "Runnable processes -4482185767840725.00" in AIX3 with xlC 2.1.3?
A: There seems to be a bug in that version of xlC. Compile the
   monitor with -O2 flag (see the Makefile).

Q: How can I display multiprocessor information in monitor?
A: Use monitor version 1.14 or 2.08 and higher with the -smp option.

Q: Is there further documentation on different monitored variables.
A: Some information is available from my conference paper on SHARE/1994
   at ftp://ftp.funet.fi/pub/unix/AIX/RS6000/monitor-paper.ps.Z
   or from http://www.csc.fi/~jmaki

Q: How can I redirect the output of monitor to a file for preprocessing?
A: Since version 2.00 you can use the -l and related options to dump
   interval (and sample) data to a logfile.
   An other way to do this interactivly is to use x-windows and copy and
   paste text region to file or editor.

Q: On IBM SP, why the css0 is showing always 0 kbyte/s transfer rates?
A: IBM/Kingston has decided that it is not worth updating the 32 bit
   network counters which are shown also by SNMP variables. The
   packet counters are however updated and thus meaningfull also with monitor.

Q: What is the process 514 which consumes much of the CPU-time?
A: It is Kernel (wait) process which is the idle loop in AIX. Wait-process
   is just the place where idle cpu-time is consumed. It is not harmfull.
   You cannot kill it. It's running in priority 127 which is the worst
   priority. Se also next question.

Q: What are all these 'kproc' processes.
A: A kernel process is one that is born in the kernel, and never leaves it.
   Sort of like a never-ending system call. There's a section in Info called
   "Using Kernel Processes" that talks about them. Various AIX subsystems use
   kprocs for maintenance tasks, such as clean ups or responses to interrupts
   that don't have to be done on-level.

   Here are a few:
    wait
       AIX idle process. This one will eat all cpu if no-one else wants it :-)
    gil 
       The "gil" kproc does timer driven TCP/IP work.  As I understand it, it 
       wakes up every 200ms and every 500ms to deal with things like packet 
       retransmissions, etc.
    swapper
       AIX pager/swapper
    jfsc
       Compressed file system process
    kbio
       Kernel biod (sort of NFS...)
    netm
    lrud

Q: I just booted my RS/6000 and why all memory is used?
A: AIX buffers all file accesses throught memory. So this is the reason
   why "all" memory seems to be used. One can look this with monitor
   in Memory part, the files shows the amount of memory that is
   used by files (actually other than text and data (procs) memory segments).
	Memory    Real     Virtual
	free      1.4 MB   42.9 MB
	procs    20.5 MB   21.1 MB
	files    10.1 MB          
	total    32.0 MB   64.0 MB

Q: What are the wierd numbers that are shown on my three digit LEDs display?
A: You are probably running older version of loadavgd or loadavgLED which 
   shows your systems load average in three digit LEDs display.

Q: Where can I find more information on how to use monitor and what
   is the meaning of all the different numbers shown by it? And how can
   I tune my system to work better?
A: At the moment there isn't any really good all explaining document on this.
   Jussi is/was working on it at his spare time.
   There is an IBM AIX publication "IBM AIX Performance tuning quide" which
   can help you solving some of the problems.

Q: How do I install monitor.
A: Assuming you are installing the binaries to /usr/local/bin
   Run as root.
       cd "directory-where-monitor-sources-are"
       make clean
       make
       ./monitor -C 1
       make install

Q: What are the max values in the logfile?
A: Take cpubusy as an example. The normal value is the averag over the interval:

        cpu_counter_start_interval  -  cpu_counter_end_interval
        -------------------------------------------------------------------
                           seconds_per_interval

   An interval contains one or more samples. For each sample the average is
   calculated in the same way (just s/interval/sample/). Now cpubusy_max is
   the maximum of these sample averages in this interval.

Q: Real memory statistics are wrong. What can I do?
A: Memory statistics are taken from an undocumented kernel structure.
   Once in a while the AIX development team changes this structure a bit,
   so monitor faults on memory stats. The monitor distribution has a small
   program that dumps this kernel structure to stdout. You can help solving
   this by running a few commands and sending me the output. Run the commands 
   in situations where tha machine is idle and very busy. Some commands may
   not be installed on your system. Try to get them from your installation 
   tapes/CDs or just forget about that part...
   The commands are in the shell script vmker.sh. This script will create a
   vmker.log.<datetime> file filled with the output of the commands ran by
   the script.  Please sent me these log files...
