Need urgent Help for Kernel upgrade......

[Former Member]

Hi Gurus,

I have recently upgraded SAP kernel from Release 6.40 Patch 109 to Release 6.40 Patch 175. My System is SAP ERP Central Component 5.0 and Release is 6.40 and Kernel Stataus is showing below details:

Kernel release 640

Compilation SunOS 5.8 G

Patch Level 175

ABAP Load 1521

CUA load 16

Mode opt

~

I have total 8 Application servers. today I have genereated EWA report and in that report, for SAP kernel section it has reported that <b>"You are running different SAP kernel releases on your instances. This situation may lead to severe problems and is not supported by SAP.

Recommendation: Use only one kernel release for all instances."</b>

Actaully I have updated kernel in my central instance (/sapmnt/<SID>/exe) and this directory is NFS mount in all the application server. So I am not sure why the SAP EWA is repoting this that you are running difference SAP Kerenl release.

If this is correct then how to rectify it ? what action I need to take to resolve this ?

Appreciate your help.

Thanks & Regards

Basis

[Former Member]

Hi Markus,

As suggested, I have checked all the application servers from SM51 -> GOTO -> Server Information -> Release notes.

For all the servers, it shows same kerenl Release and patch. The details is as given below:

~

SAP R/3 Kernel information

SAP R/3 Kernel 640_REL

Database Client Library OCI_920

created in SunOS 5.8 Generic_117350-38 sun4u

created on Mar 21 2007 21:10:36

Support Level 0

Kernel Patch number 175

Source ID 0.175

DBSL Patch No. 175

ICU Version 2.6.1 Unicode Version 4.0

~

I am still fail to understand, from where it is getting the information having difference kernel for difference instance.

Also I have checked SM51 -> GOTO -> SAP Directories and found follwing:

DIR_CT_RUN /usr/sap/PRD/SYS/exe/ctrun

DIR_EXECUTABLE /usr/sap/PRD/SYS/exe/run

DIR_EXE_ROOT /usr/sap/PRD/SYS/exe

The above is discussed in the SAP note. So not sure what is wrong where. ?

any guesses...?

Thanks & Regards

Basis

[markus_doehr2]

you have the (old) "ctrun-mechanism" already set - so everything is correct.

mabye you libdbora.so is diferent? If you can´t find an error here, the EWA may be wrong.

--

Markus

[Former Member]

Hi Markus,

I will plan to do that as early as possible. Even I am going to write to SAP to release that note in english for better understanding.

My question is: Am I on HIGH ALERT for this issue ? Since it is production server, will it going to affect my system in any aspect ??

2nd thing, This is applicable only to NW ? Because I have upgraded SAP Kernel to Patch 175 to other R/3 system having more application server, but I didn't get this comment in my EWA report ?

This week I am planning to upgrade kernel for 4.6D release to Patch 2307. Will this be gona issue for that also ?

Please advise.

Thanks & Regards,

Basis

[markus_doehr2]

I will plan to do that as early as possible. Even I

am going to write to SAP to release that note in

english for better understanding.

yes - that note should be translated giving the importance.

My question is: Am I on HIGH ALERT for this issue ?

Since it is production server, will it going to

affect my system in any aspect ??

No, you will just see a slightly higher usage of "nfsd" when you start the application servers for the first time since they will all copy parts of the kernel locally.

2nd thing, This is applicable only to NW ? Because I

have upgraded SAP Kernel to Patch 175 to other R/3

system having more application server, but I didn't

get this comment in my EWA report ?

The comment isn´t related to to usage of sapcpe. The comment means, that you have at least one application server, that uses local binaries with a different patchlevel instead of those on the central instance. I would consider that more critical. Maybe you installed one of the servers later and used a newer installation CD set which creates that new structure for using sapcpe by default... don´t know, just a guess. To check, go to SM51 and select each server and press on "Release Information", you will see which one is affected.

This week I am planning to upgrade kernel for 4.6D

release to Patch 2307. Will this be gona issue for

that also ?

Kernels lower than 6.10 use a different approach (the "ctrun directory" mentioned in that note) - technically it´s the same. The installation guides for 4.6c should tell you how to set that up.

--

Markus

[Former Member]

Hi Markus,

Appreciate your prompt reply. I have checked the note but it is in German language. Do you know the english version?

I am not aware of this ctrun concept. What do I do now ? it will gona impact my sytem badly ? how to resolve it...

Thanks & regards

Basis

[markus_doehr2]

I saw that, the note is not (yet) translated - or won´t be - don´t know.

I can try to explain

Early releases of NFS were pretty buggy, especially those running over UDP, you could have had stale handles or strange signals on the system, that mounted the filesystems over NFS.

To make execution of kernels with growing sizes faster, SAP "invented" sapcpe, which is a tool, that copies necessary parts of the kernel locally to the application server. With that approach you have two advantages:

- copies of the kernel can be (re-)started faster

- consistent versions of the kernel on all application servers in that instance

- patching involves only the installation of one installation on the central instance

To use that mechanism (using that note)

- shutdown all SAP instances on all servers

- create /usr/sap//exe explicitly on the application server

- change the profile parameters for all systems in all profiles according to the note

(a lot of work - but surely worth it if you have 8 application servers)

--

Markus

[Former Member]

Just to add the unix filesystem defined on application server.

~

/usr/sap/PRD/SYS/exe

~

lrwxrwxrwx 1 root other 15 Mar 22 2005 dbg -> /sapmnt/PRD/exe

lrwxrwxrwx 1 root other 3 Mar 22 2005 run -> dbg

drwxr-xr-x 2 prdadm sapsys 512 Jun 1 2004 opt

[markus_doehr2]

Check

Note 1104735 - Upgrade to the new Instance-Specific Directory on UNIX

You doesn´t seem to use the "ctrun" approach mentioned in that note.

--

Markus