I often assist people who are new to IBM I and sometimes
they lack some basic knowledge that would really help them in their day to day
work.
I thought I would write a blog in which I will document some
of the more common issues which occur.
First, I will point out some extremely simple performance
tasks that will catch many potential general performance problems.
First, we have WRKDSKSTS.
This displays a list of all your disk arms, and how much
they are being accessed.
For performance issues, we are most interested in the “%
Busy” column. In order to analyse correctly, you must go to this screen, let it
run for 20 minutes or so, then press f5 to refresh.
If a disk arm is busy for approximately 20% or more, this is
a bottleneck and a potential cause of a problem. Over 30% can indicate a
critical issue.
Resolving this issue involves running TRCASPBAL as described
in SAP note 517515. The classic cause for this issue is the installation of a new
disk arm leading to an uneven distribution of data.
Second, we have WRKACTJOB.
This displays all the active jobs on the machine, divided
into the various subsystems, including one for each instance (usually named
R3_<NN>, NN being the instance number) and information about them.
The column we are interested in here is “CPU %”. Select it,
then select f16 (alt f4). Again, f5 will refresh the information for you.
This will organise all the jobs running into order of CPU
consumption. If you see one which is conspicuously high, then select 5 in the “Opt”
column. This will allow you access to the job log of the job (option 10), work
with its locks (option 12), or even end the job if you are sure it is safe to
do so.
Finally, we have WRKSYSSTS. This is the most useful of the
three screens for telling the overall health of your system.
There are several areas we are interested here.
First is % CPU consumption. This will show the overall
consumption for the partition. If, for example, you had 4 CPUs, and the figure
was 20%, 4 CPUs could be at 20% each, or 2 could be at 40%. It should be noted,
however, if you are using a partitioned system with uncapped partitions you may
see a figure over 100% here, so do not panic!
The second thing you need to check is the % of the system
ASP used. If your system ASP nears being full you will start to face many
issues. The default for IBM i will present warnings for this if this breaches
90%.
You can also see from this screen if the issue is temporary
storage (Current unprotect used) which can be examined with DSPTMPSTG.
The final thing I would like to discuss is page faulting. In
the bottom half of WRKSYSSTS is a section named “System Pool”. The pool we are
interested in is pool 1- also known as the machine pool.
If you see a fault rate of over 10 here, it is a sign that
the memory assigned to the machine pool is too small. It is a simple matter to
change this though as you can easily reassign memory from WRKSYSSTS itself in
the Pool Size field.