topic Re: m/cs hangs if the number of user increases. in Operating System - HP-UX

m/cs hangs if the number of user increases.

Sanjiv Sharma_1 — Mon, 10 Jul 2000 09:03:12 GMT

Dear Friends,

I am facing a peculier problem the details of which are given below :
prob: when the no. of users becomes > 50 the m/c becomes very slow and when it becomes around 100, it almost hangs.

configuration: HP E55 with 10.20 and oracle 8.0.4
software: HPUX unlimited userlicence.
C/ANSI C Compiler
XSW800GR1020
XSW800HWCR1020

value ofkernel parameters:
maxuprc 500
nproc 4116
maxusers 512
nbuf 18120
npty 715

Pls. help me to resolve this.

Re: m/cs hangs if the number of user increases.

CHRIS_ANORUO — Mon, 10 Jul 2000 09:15:47 GMT

Hi Sanjiv,
You have to retune your kernel with the following values:
nbuf=0
bufpages=0. This will enable your system to do dynamic memory paging within the set values of dbc_max_pct and dbc_min_pct.
maxuproc=nproc-4
allocate_fs_swapmap=0
swapmem_on=1
Reduce the values of maxusers and ntpy to 250 since the number of users is at 100 max.
ntpy is for number of telnet connections, and maxusers is a macro.

Re: m/cs hangs if the number of user increases.

Tim Malnati — Mon, 10 Jul 2000 12:59:31 GMT

The kernel ajustments noted by others may help, but you may be at a point where your asking too much of that little box. I suggest you start using Glance and sar to get an idea of what's going on. If you have not purchased Glance, there is a trial version available at software.hp.com

Re: m/cs hangs if the number of user increases.

Victor BERRIDGE — Mon, 10 Jul 2000 16:50:48 GMT

Hello Sanjiv,
There is one thing you havent mentionned: How much RAM do you have on your E55.
As Tim mentionned you are maybe asking too much for its configuration.
Have a look at swapinfo.
Is there a reason to give maxuprc such a high value, for me unless a user comes around and proves me he needs more I set the value at 75 (and progressively increase to 1000 if I have lots of RAM). That is greater this value,slower is your machine when thing dont go right:
sql request badly written, higher timeout in loops etc...
I would agree with Chris set your values to a smaller configuration and on request increase the values where needed.
I know that its a small machine but well tuned it should support with ease 100 concurrent users with oracle and lots of disks (mine do).
Heres one of my machine parameters:
maxuprc 75
bufpages 131072
nfile 1610
nbuf 8832
nproc 532
npty 148
timeslice 10

Feb 1 08:53:21 regulus vmunix: 9245XB HP-UX (B.10.20) #1: Sun Jun 9 06:31:19 PDT 199
6
Feb 1 08:53:21 regulus vmunix:
Feb 1 08:53:21 regulus vmunix: Memory Information:
Feb 1 08:53:21 regulus vmunix: physical page size = 4096 bytes, logical page size = 4
096 bytes
Feb 1 08:53:21 regulus vmunix: Physical: 131072 Kbytes, lockable: 91584 Kbytes, avail
able: 108044 Kbytes
# model
9000/856/E55
# uname -a
HP-UX regulus B.10.20 C 9000/856 149934332 32-user license
#
# swapinfo
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 262144 0 262144 0% 0 - 1 /dev/vg00/lvol2
dev 524288 20540 503748 4% 0 - 0 /dev/r5vg02/lvol2
reserve - 47648 -47648
memory 92508 68860 23648 74%
#
I hope It will help a bit

Best regards
Victor