Sun V880 Solaris 8 - 8GB RAM

I have a bit of a strange scenario and I have also logged this with
Oracle just in case it's an Oracle problem.

We have installed the oracle software under separate homes and there
are 2 users running a number of databases

- oracle
- devdba

1 - The databases under the oracle user have no problems.

2 - The database under devdba crash occasionally.
- Their background processes appear to die.
- The oracle memory structures are in place and we can query this.
- We have to restart the devdba databases to do any work
- There is no errors written to any oracle trace file.

3 - There is nearly double the amount of devdba databases than there
is oracle.

4 - Are we hitting some unix limit restriction for the devdba user -
is this possible?

I was checking the parameter maxusers as it sets the limit on the
number of processes running for a user.

From /etc/system
* To set variables in 'unix':
*
* set nautopush=32
* set maxusers=40

The number of processes currently running on the box are:

pfstst1:IBS:/var/adm> ps -eaf|awk '{print $1}'|grep oracle |wc -l
50
pfstst1:IBS:/var/adm> ps -eaf|awk '{print $1}'|grep devdba |wc -l
178
pfstst1:IBS:/var/adm> ps -eaf|awk '{print $1}'|grep root |wc -l
55

The only error message in /var/adm/meesages is below.

Any ideas? Is it a threshold limit of some sort?


Feb 25 18:25:30 pfstst1 dtexec[1110]: [ID 172582 user.error]
libtt[1110]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 18:32:22 pfstst1 dtexec[1689]: [ID 508517 user.error]
libtt[1689]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 18:32:22 pfstst1 dtaction[1690]: [ID 306214 user.error]
libtt[1690]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 18:32:23 pfstst1 dtexec[1692]: [ID 454809 user.error]
libtt[1692]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 19:55:54 pfstst1 sudo: [ID 702911 local2.alert] irwinme :
command not allowed ; TTY=pts/6 ; PWD=/u01/home/irwinme ; USER=root ;
COMMAND=/usr/bin/su -devdba
Feb 25 19:59:18 pfstst1 dtexec[7507]: [ID 228434 user.error]
libtt[7507]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 19:59:23 pfstst1 dtexec[7517]: [ID 244818 user.error]
libtt[7517]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 19:59:23 pfstst1 dtaction[7518]: [ID 769106 user.error]
libtt[7518]: _Tt_session::client_session_init(): ! Tt_status
Feb 25 19:59:23 pfstst1 dtexec[7520]: [ID 191110 user.error]
libtt[7520]: _Tt_session::client_session_init(): ! Tt_status


Thanks for your time.

cheers,
Biagio

bidge wrote:
> Sun V880 Solaris 8 - 8GB RAM
>
> I have a bit of a strange scenario and I have also logged this with
> Oracle just in case it's an Oracle problem.
>
> We have installed the oracle software under separate homes and there
> are 2 users running a number of databases
>
> - oracle
> - devdba
>
> 1 - The databases under the oracle user have no problems.
>
> 2 - The database under devdba crash occasionally.
> - Their background processes appear to die.
> - The oracle memory structures are in place and we can query this.
> - We have to restart the devdba databases to do any work
> - There is no errors written to any oracle trace file.

Sounds a bit of misconfiguration of the devdba databases. Are the
databases usually started in a particular order? If so, try
reversing that and see what happens.

> 3 - There is nearly double the amount of devdba databases than there
> is oracle.
>
> 4 - Are we hitting some unix limit restriction for the devdba user -
> is this possible?
>
> I was checking the parameter maxusers as it sets the limit on the
> number of processes running for a user.
>
[cut]

Well, per your subject line, Solaris does not kill processes automatically.
You can configure resource management to limit the number of processes
a user may have, number of lwps in total etc. The 'maxusers' is awfully
named - it as nothing to do with any maximum user limit, it is just
used for sizing a number of tables and caches; 'maxusers' is set
based on physical memory installed.

You need to gather some data. Choose a devdba process that is likely to
die and truss it to see it's death throes. By 'background processes' do
you mean the core oracle processes such a log writer, db writer etc or
user-initiated jobs?

Gavin

You may need to increase the file descriptor size for devdba.

Mark

Gavin Maltby <G_a_v_i_n.M_a_l_t_b_y@sun.com> wrote in message news:<c1kgd3$k9n$1@new-usenet.uk.sun.com>...
> bidge wrote:
> > Sun V880 Solaris 8 - 8GB RAM
> >
> > I have a bit of a strange scenario and I have also logged this with
> > Oracle just in case it's an Oracle problem.
> >
> > We have installed the oracle software under separate homes and there
> > are 2 users running a number of databases
> >
> > - oracle
> > - devdba
> >
> > 1 - The databases under the oracle user have no problems.
> >
> > 2 - The database under devdba crash occasionally.
> > - Their background processes appear to die.
> > - The oracle memory structures are in place and we can query this.
> > - We have to restart the devdba databases to do any work
> > - There is no errors written to any oracle trace file.
>
> Sounds a bit of misconfiguration of the devdba databases. Are the
> databases usually started in a particular order? If so, try
> reversing that and see what happens.
>

The databases are not started in any particular order.

> > 3 - There is nearly double the amount of devdba databases than there
> > is oracle.
> >
> > 4 - Are we hitting some unix limit restriction for the devdba user -
> > is this possible?
> >
> > I was checking the parameter maxusers as it sets the limit on the
> > number of processes running for a user.
> >
> [cut]
>
> Well, per your subject line, Solaris does not kill processes automatically.
> You can configure resource management to limit the number of processes
> a user may have, number of lwps in total etc. The 'maxusers' is awfully
> named - it as nothing to do with any maximum user limit, it is just
> used for sizing a number of tables and caches; 'maxusers' is set
> based on physical memory installed.
>
> You need to gather some data. Choose a devdba process that is likely to
> die and truss it to see it's death throes. By 'background processes' do
> you mean the core oracle processes such a log writer, db writer etc or
> user-initiated jobs?
>
> Gavin

Thanks Gavin.
I do mean the core oracle processes as you mentioned. I can actually
log into the db and query the v$session view. It returns all session
details including the oracle background processes. When I go to unix
prompt and grep the oracle processes it returns the user connections
but not the oracle background processes.

1 select SID,USERNAME,PROCESS
2* from V$session
SQL> /

SID USERNAME PROCESS
---------- ------------------------------ ---------
1 8928
2 8930
3 8932
4 8934
5 8936
6 8938
7
8
9 PFS_USER 1424:1220
10 PFS_USER 1624:1400
12 PFS_USER 1424:1940
13 PFS_USER 1424:1860
14 PFS_USER 1424:2128
15 PFS_USER 1424:2084
16 PFS_USER 1424:496
17 PFS_USER 1424:1220
18 PFS_USER 1424:1456
19 SYS 29431
22 ORACLE 2903

I will truss one of the database processes and if it dies again
hopefully this will help.

Thanks.

mdhmi@yahoo.com (Mark Daniels) wrote in message news:<7ee5b960.0402260740.59fd9d15@posting.google. com>...
> You may need to increase the file descriptor size for devdba.
>
> Mark

How do I do that?
Thx.