Re: HP-UX

  • From: "LS Cheng" <exriscer@xxxxxxxxx>
  • To: aleon68@xxxxxxxxx
  • Date: Tue, 15 Aug 2006 22:32:17 +0200

Hi

If most of your operations are tiny I/O operations such as index scans then
your SAN statistics would not show anything that could alert you and your
SAN Administrator. I had a similar problem last year with a EMC DMX San,
when I went to ask the SAN Admini he was showing me how good was SAN's
buffer hit ratio and how much bandwidth the SAN still has (was only 50% busy
according to him).

Not sure if you have enabled the read cache in your HDS SAN?

If you are using Online JFS you would use the tipical parameters such as
convosync = direct, mincache = direct, nodatainlog, delaylog


On 8/14/06, Alfonso León <aleon68@xxxxxxxxx> wrote: 

 Hi Every one:
we went online today with full load, until last week we were 20% loaded,
and CPU Utilization  with less 10%. We have a HPUX with 10 cpus and a
Hitachi SAN. Oracle 9.2.0.6. CPU utilization is now 40%.

Today the performance was way more less than expected.

these are the top wait event


Top 5 Timed Events

~~~~~~~~~~~~~~~~~~ % Total

Event Waits Time (s) Ela Time

-------------------------------------------- ------------ -----------
--------

db file sequential read                      1,768,294       38,259
39.69

log file sync                                    71,190      30,065
31.19

free buffer waits                                12,212      11,351
11.78

buffer busy waits                               360,596       8,773
9.10

CPU time                                                      2,736
2.84

db file scattered read                          382,133       1,691
1.75

db file parallel read                             2,958       1,199
1.24

log file parallel write                           5,193       1,073
1.11

enqueue                                           1,649         698
.72

SQL*Net message from dblink                      54,991         141
.15

As you can see i'm waiting on disk, the path utilization of the san is
less than 20% and the CPU utilization is less than 40%. almost all the
processes are sleeping.

On the HP-UX monitor says that peak disk utilization is 100% the rest of
the parameters are normal and sub-used.

Does anybody know if there is parameter in kernel that limit the disk
request?

Any idea would be appreciated.


Alfonso Leon

  • References:
    • HP-UX
      • From: Alfonso León

Other related posts:

  1. Home
  2. oracle-l
  3. Archive
  4. 08-2006