system disks on aix 5.3

UNIX AIX
1

hello

i'm running on P570 box aix 5.3 8 cpus 24G ram
there are 1850 users loged in to this box

the problem is that the two sysytem disks busy all the time

hdisk0 100% busy
hdisk1 100% busy

some one have an idea what writing to this disks?

thanks

ariec

2

Why don't you have a look? Monitor the disk activity, there are plenty of tools for that purpose: iostat, nmon, topas, ....

If you want an in-depth analysis about what exactly is going on in your filesystem use "filemon" over 5-10 seconds and analyze its output. Read carefully what the manpage says about filemon, as it is not trivial.

Lacking any further information i would bet my money on swapping though: do an "lsps -a" and tell us what the figures are. Also the output of "svmon -G" would be helpful.

I hope this helps.

bakunin

3

hello

first thanks for your help

second all the details that i've wrote comes from topas iostat and all the tools that you wrote.
my pagesapce is on emc disks, out of the machine.
on system disks i've only filesystems system
can you have more ideas?

thanks

ariec

4

filemon and RPM lsof

5

where i can see jobs who send from unix and did not get out to printer?

6

hello

i've typed who -r command on aix 5.3 box and no answer
how do i know on what run level my machine P570 is running?

thanks

ariec

7

o14: /home/vbe> oslevel
5.3.0.0
o14: /home/vbe> who -r
. run-level 2 Feb 08 14:13 2 0 S

8

Until now you haven't posted any "details", you just told us that the disk activity is high. If you have used "all the tools" i mentioned you should know already by now the reason for this and don't need any advice.

bakunin

9

hello

i'm have P570 24Gram 8 CPUS AIX 5.3
in mysterious way and very sad surprise in the midle of the day when 1700 users loged in to the machine i've found a lot of defunct proccess.
the number of defunct bocome bigger and bigger 53000 proccess.
the machine become hang.
no changes in system.
the machine run application + Ingres 2006 DB.
one application who all the users do login and run this application.
there are 9 instalaions of Ingres sperates.

some one have an idea what can be the cause of this case

thanks

ariec

10

A Zombie (a "defunct" process) is not the fault of the OS it is the fault of sloppy application programming. Without further analysis of what the 1700 users are doing on the machine your problem cannot be solved.

Anyways, 1700 users are an awful lot, if they are all connected via regular login sessions. I would *expect* problems with such a heavy load. UNIX systems are midrange systems historically and when it comes to sheer user numbers this is showing. A host system (s/390, z/OS, etc.) would handle such amounts of user sessions with ease, but a UNIX system is simply not built to cope with such numbers. I'd suggest splitting the load between several login servers (create several smaller LPARs instead of the single big one) and use an application server behind these.

I hope this helps.

bakunin