Error while installing NetWeaver 7.0 double stack with PI

[Former Member]

Hi,

we're installing NetWeaver 7.0 Java+ABAP with PI under Solaris 10 with an Oracle DB. The server has 16GB of RAM.

In phase 21 (Start instance) we get the following error:

ERROR 2007-08-15 13:18:55

CJS-30149 ABAP processes of instance <sid>/DVEBMGS86 [ABAP: UNKNOWN] did not start after 10:30 minutes. Giving up.

ERROR 2007-08-15 13:18:55

FCO-00011 The step start with step key |NW_Onehost|ind|ind|ind|ind|0|0|NW_Onehost_System|ind|ind|ind|ind|1|0|NW_CI_Instance|ind|ind|ind|ind|11|0|NW_CI_Insta

nce_Start|ind|ind|ind|ind|1|0|start was executed with status ERROR .

When trying to start the instance manually, it says "instance already running".

Any help would be highly appreciated.

Thanks,

Paul

[markus_doehr2 · Accepted solution]

What do you see in the workprocess traces (dev_disp and dev_w0) in

/usr/sap/<SID>/DVEBMGS86/work

and hints there why the system is not coming up?

--

Markus

[Former Member · Accepted solution]

Markus,

we don't have a dev_w0 but dev_disp contains the following error lines:

• ERROR => e=28 semget(28632,1,2047) (28: No space left on device) [semux.c 924]

• ERROR => e=28 semget(28632,1,2047) (28: No space left on device) [semux.c 924]

• ERROR => SemRq: Implicit SemInit failed. Key=32 [semux.c 1544]

• ERROR => SemRel: Ill. internal Handle. Key=32 [semux.c 1726]

<ES> client 0 initializing ....

• ERROR => e=28 semget(28633,1,2047) (28: No space left on device) [semux.c 924]

ThSendErrMsg: called during initialization

in_ThErrHandle: 1

• ERROR => ThSendErrMsg: db_connect (step 1, th_errno 13, action 3, level 1) [thxxhead.c 10156]

Any ideas?

Paul

[markus_doehr2 · Accepted solution]

So you shared memory parameters are not configured.

Did you create a project file (/etc/project) for this instance? On Solaris 10 you don´t use any more /etc/system to set those parameters...

--

Markus

[Former Member · Accepted solution]

Markus,

thanks for your quick reply. There is a /etc/project and it contains the line .max-shm-memory=(priv,18446744073709551615,deny).

By the way, the instance we are installing is the second one on the host and it is supposed to share the DB with the other one.

Paul

[markus_doehr2 · Accepted solution]

And the user of that second instance <sid>adm is in the /etc/user_attr? Otherwise the project files settings will not be taken:

/etc/user_attr:

<sid>adm::::project=<yourproject>

--

Markus

[Former Member · Accepted solution]

Markus,

the file /etc/user_attr did not contain the perspective line. I added the line which then seems to cause sapinst to freeze in the same phase (21). In the status line it says "Stopping instance <sid>/DVEBMGS86...". The file dev_disp doesn't exist yet at this point.

Also, we're unable to stop sapinst. Even kill doesn't do it - only a restart. This happened twice now.

Any idea why this is?

Paul

[markus_doehr2 · Accepted solution]

The problem seems to be, that the already allocated shared memory segments for the new instance are not yet released.

Check Note 724713 - parameter settings for Solaris 10 if all the necessary parameters in your project file are there.

What version of Solaris 10 do you use? (cat /etc/release)

--

Markus

[Former Member · Accepted solution]

That helped!

Thanks a lot!

Paul

[Former Member · Accepted solution]

Unfortunately, there's another problem a few phases later. In phase "Install Java Engine" sapinst just hangs. The status bar says "Inserting Administrator credentials in secure store".

I found different suggestions on the web. I tried editing a line in <install_dir>/control.xml, I made sure JAVA_HOME is set correctly and is also in the PATH, I retried several times, no luck...

Any more great ideas?

Paul

[markus_doehr2 · Accepted solution]

What JDK version are you using?

--

Markus

[Former Member · Accepted solution]

We have 1.4.2_11-b06.

Paul.

[markus_doehr2 · Accepted solution]

I would recommend using 1.4.2_12 at least, better 1.4.2_14 or_15.

Do you have the JCE policy file installed? Is the machine doing something? Check with

prstat -a 1

if you see any CPU usage...

--

Markus