topic Re: 100 % Disk Utilisation in Glance in Operating System - HP-UX

100 % Disk Utilisation in Glance

yc_2 — Thu, 05 Jul 2001 07:43:39 GMT

Hi,

My system (N-440, HPUX 11.00, INFORMIX RDBMS) shows 100% Disk utilisation in glance and memory page shows:

Page Faults 934 209077
Page In 362 82756
Page Out 3 730
KB Paged In 0kb 152kb
KB Paged Out 12kb 2.9mb

swapinfo -tam
Mb Mb Mb PCT
TYPE AVAIL USED FREE USED dev 4096 0 4096 0% dev 17366 0 17366 0% reserve - 1256 -1256
total 21462 1256 20206 6%

Could it be the kernel parameters setting that caused the high disk utilisation? If so, which are the parameters? If not, please advice.

Thanks in advance,
YC

Re: 100 % Disk Utilisation in Glance

Bill McNAMARA_1 — Thu, 05 Jul 2001 07:48:30 GMT

which disk device file is the major culprit identified by glance and/or iostat -t 5?

If it is your swap device disk then yes, memory tuning will help a bit.. but a ram upgrade would be better!

If it is just standard usage to your db, then you've got to investigate lvm configuration, effective load balancing, disk array configuration and so on.

ioscan -fnkCdisk
vgdisplay -v vgdb
mstm tools-info-run on the disk array component

Later,
Bill

Re: 100 % Disk Utilisation in Glance

Ravi_8 — Thu, 05 Jul 2001 07:59:11 GMT

Hi,
additional RAM solves this problem. by looking at glance you can't say 100% disk is full use 'bdf' command, it gives the full dteails.

Re: 100 % Disk Utilisation in Glance

Steve Lewis — Thu, 05 Jul 2001 11:55:09 GMT

The kernel parameters themselves may affect the DU, depending on how your system is set-up. Maybe if you posted your system file and onconfig, onstat -d, onstat -g iof, onstat -g seg, vmstat -S -n, etc we could all provide better advice. Tell us more about the model of system, the RAM, CPUs and apps.
It would be odd indeed if your running kernel took all the memory. Warning: some of the informix recommended parameters (e.g. NFILE, SEMUME) are wide of the mark and go beyond what is necessary for a well tuned system, unless your installation is huge. Other problems people sometimes have include shmmax and dbc_max_pct. Shmmax is not the size of a segment but the maximum permissible size. Onconfig parameters SHMVIRTSIZE and BUFFERS (use onstat -g seg and ipcs -ma) determine the actual size of shared memory segments. It may be possible to reduce these a bit and still get reasonable performance. But obviously the more RAM you have, the better.
You may also get high i/o problems due to application issues.