Operating System - HP-UX
1753672 Members
5654 Online
108799 Solutions
New Discussion юеВ

STRMS stop reason - GlancePlus

 
LudekC
Advisor

STRMS stop reason - GlancePlus

Good day,
configuration:
HP-UX B.11.11 not patched
application server rp5470 (4 CPU)
database server rp2470 (2 CPU)- Oracle 9.2.0.4
connection betwenn servers - network 1GB

Database server - all database process oracleagro are in STRMS stop reasn

Application server - all user proces are in STRMS stop reason, some in PRI.

Is this problem? How determine where is problem? How increase server performance with data on Oracle?

Thank You Ludek
4 REPLIES 4
John Guster
Trusted Contributor

Re: STRMS stop reason - GlancePlus

"wait reason is streamwait" means wait for an operation to complete while accessing a "stream" device. It is normal stop reason for kernel threads and daemons waiting for a stream event which includes the network transport and pseudo ternimal I/O requests.
So it is normal.
Bill Hassell
Honored Contributor

Re: STRMS stop reason - GlancePlus

The label in Glance is very misleading...it does not mean that the program is stopped. It means that during the extremely small measurement period, the program had initiated a task which requires waiting until the task is complete. It makes no sense for a program to use data before it is ready. Whether the wait is for I/O, a semaphore, a streams lock, etc, this is perfectly normal and has nothing to do with performance of your system.


Bill Hassell, sysadmin
LudekC
Advisor

Re: STRMS stop reason - GlancePlus

Good day,
I watch in GlancePlus network:
NET2: between application - database server for example Packet in rate = 3000 a Packet out rate = 2100. How determine rate in MB?
NET1: application server - user network for example Packet in rate = 120 a Packet out rate = 70.
Is it normal?
Thank You Ludek
Bill Hassell
Honored Contributor

Re: STRMS stop reason - GlancePlus

You wrote:

>NET2: between application - database server for example Packet in rate = 3000 a Packet out rate = 2100. How determine rate in MB?

> NET1: application server - user network for example Packet in rate = 120 a Packet out rate = 70. Is it normal?

The rate in MB cannot be determined because tyhe packet size is variable (1 byte to 1500 bytes apx). 3000 packets per sec is fairly normal but networking is extremely tricky as a metric because it depends on both ends. A very fast server will look very slow if the other side of the connection is very slow. A 1Gb network connection will not perform at full speed when talking to a 10 Mb PC LAN card. Since it appears that the bulk of the traffic is between the app and database servers, that appears to be optimal.

> How increase server performance with data on Oracle?

There is no simple answer at all. The first step is PATCH!!! You cannot expect good performance (or reliability for that matter) without applying the current patch set. The next step is to analyze your Oracle stats (you do have StatsPak, correct?). The numbers may tell you that you have crippled your Oracle engine with far too little memory and may have to add 4-8Gb more RAM and increasing SGA significantly. Or the stats for your slowest SQL task show that there are missing or unbalanced indexes.

The vmstat command will tell you if RAM is too small: look at the po column. 2 digits (ie, 10-99) and especially 3 digits or more indicate a serious memory shortage. This will slow all processes to a crawl.


Bill Hassell, sysadmin