cancel
Showing results for 
Search instead for 
Did you mean: 

ECC 6.0 installation on Sybase and AIX 7.1 using Power HA

Former Member
0 Kudos

Hello Experts,

We have a requirement wherein we need to install ECC 6.0 EHP 6 Production system on Sybase and AIX 7.1 and using Power HA(earlier known as HACMP).

We are required to install database on let's say HOST A, and CI on HOST B. Both HOST A and HOST B are in a mutual fail over cluster,as prepared by hardware vendor.

Being new to High availability installations on SAP, we would initially want to know if for this type of installation we need to go for HA option in sapinst/SWPM or it should be the distributed system installation, on different hosts?

In HA installation we have an ERS (enqueue replication server) installation option which we do not want to install,so please let us know if in HA installation option, ERS is a mandatory installation.

Any suggestions for file systems(mount points,shares) in database host and CI would be highly appreciated.

Just to add that we have gone through installation guides and many other SCN and help.sap.com links, but could not find concrete information and got further confused.

Your help and guidance in this regard will be highly appreciated.

Regards,

Bharat

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Hi Bharat,

              You can definitely have below scenario without ERS server.

Node A : DB+CI

Node B:  ASCS

Node C:  APP

               

             Where Node A and Node B are in cluster and Node C is not in cluster.

             Your SPOF are DB and ASCS.

              Here even you CI is behaving as a normal APP server.

              Since your ASCS includes Message server and Enqueue server(Lock Table)

              it is not mandatory to have ERS as part of cluster but it is SAP best practice

             to install ERS as if something happens to your lock table in ASCS instance then

              you have its replica to retain back the lock entries.

              So your cluster should look like below as per deepak's comment.

Node A : DB+CI+ERS

Node B : ASCS

Node C: APP

Thanks and Regards,

Kunal Gahlot

Former Member
0 Kudos

Hi All,

Thanks alot Deepak,Roman and Kunal for your valuable inputs.They have been really helpful so far.

Another question in this regard would be the concept of virtual hostname,Do we need to start sapinst for DB and CI installation using parameter SAPINST_USE_HOSTNAME even if we are installing these on different clustered nodes.

is it possible to give a gist of why and how do we use virtual host name concept in HA environment.

Now we have started our installation as

NODEB-Installed ASCS(planning to install CI as well )

NODEA-installing DB(started without using  SAPINST_USE_HOSTNAME)

Please let me know if this installation approach is ok and should i reinstall DB using SAPINST_USE_HOSTNAME.?

Thanks a lot once again for all your help everyone.

Regards,

Bharat Pasricha

former_member188883
Active Contributor
0 Kudos

Hi Bharat,

For the SPOF services like DB , ASCS you need to use SAPINST_USE_HOSTNAME parameter while installation. This is to ensure that all the environment parameters as well as profile parameters have virtual hostname and not physical hostname.

During failover virtual hostnames gets trasferred as a part of cluster and this shall be useful in making the environment up an running on another node.

Hope this helps.

Regards,

Deepak Kori

Former Member
0 Kudos

Hello Experts,

Thanks a lot once again for all your help and guidance.We have now installed our system as following:

NODE A=ERS+DB

NODE B=ASCS+CI

/usr/sap/trans and /sapmnt/SID have been shared and mounted from NODE B to NODE A.

ASCS and DB installatiion have been done using the parameter SAPINST_USE_HOSTNAME.

Now we are installing Application servers on NODE C and NODE D for which we have only shared the intially installed  /usr/sap/trans directory with these application servers.Can anyone in a gist just tell us of what points we need to take care of while installing Dialog instances on NODE C and NODE D.

Just to confirm,NODE A and NODE B are in HACMP,

NODE C and NODE D are not part of any cluster.

Also are basis guys responsible for creating failover scripts?In order to configure HACMP between nodes, what are the tasks expected from a basis person?If anyone can provide us with a template regarding scripts,it will be very helpful and will help us in better understanding.

Thanks again everyone for helping us out in our time of need.

Regards,

Bharat Pasricha

former_member188883
Active Contributor
0 Kudos

Hi Bharat,

H/W vendors are the owners of the failover scripts. For your case you can contact IBM support to configure the failover scripts.

Hope this helps.

Regards,

Deepak Kori

Former Member
0 Kudos

Hi,

You need to share /sapmt/<SID> file system also.

HACMP scripts-check with ur AIX guys

Regards.

Ruchit

Answers (3)

Answers (3)

Former Member
0 Kudos

Hi All,

Thank you Deepak and Ruchit for your valuable inputs and suggestions.

We now have APP servers ready on Node C and Node D,and we are in the process of patching the system.We are planning kernel upgrade and will be copying the new kernel in /sapmnt/SID/exe file on CI/ASCS instance(NODE B).

and for that, what i understand i should do is

1.run stopsap command on Node C and D(Application Severs)

2.run stopsap on Node B(CI) and proceed to upgrade kernel and adjusting permissions(after taking kernel backup).

3.run startsap command in the reverse order i.e Node B,D and C.

Please let me know if the above mentioned steps are ok to proceed.Also does kernel upgrade require individual and manual kernel replacement on CI and both app servers?

Now since CI Node(Node B) is a part of cluster,do i need to make any changes in cluster settings before stopping SAP?

Also we have SAP license(Standard and Maintenanace) applied on NODE B(CI+ASCS) of the cluster via slicense,but have not applied license to other cluster node NODE A(DB+ERS),so the only way to do that is would be a manual switchover?

Thanks again everyone for your help and guidance.

Regards,

Bharat Pasricha

Former Member
0 Kudos

Hi Bharat,

The steps are okay.

Kernel upgrade doesn't have to be manual if sacpe steps are defined in start profile of the app servers since the start of SAP instance would take care of replication in case it finds any difference in the relevant binaries.

Former Member
0 Kudos

>> In HA installation we have an ERS (enqueue replication server) installation option which we do not want to install,so please let us know if in HA installation option, ERS is a mandatory installation.

Yes, you must to install ERS instance if you want to have HA. Also you need to configure ERS as cluster resource and adjust anti-collocation option. If you don't want to setup ERS as cluster resource please read about polling concept.

High Availability with the Standalone Enqueue Server

former_member188883
Active Contributor
0 Kudos

Hi Bharat,

For HA setup on AIX with Sybase you can follow  this

1) Node A: DB instance + ERS + CI
2) Node B: ASCS +APP

DB instance and ASCS instances should be part HACMP.

For Files systems
1) Include all the DB file systems in cluster
2) Include /sapmnt & /usr/sap/trans in cluster

Rest other file systems can be kept outside cluster.

Some basic information on file system based on Sybase ASE can be found in my blog which is
based on HPUX.
http://scn.sap.com/community/sybase-ase/blog/2013/08/06/things-to-consider-when-you-plan-to-install-...

Additionally refer Installation guide , section HA setup.


Hope this helps.

Regards,
Deepak Kori

Former Member
0 Kudos

Hello Deepak,

Thanks a lot for the reply.I am glad it helped me realize that i am on the right track and I have understood a few things correctly,if not all.

Our setup requires us to install DB on HOST A and CI on HOST B and then APP server on HOST C.Host A and B are clustered,and HOST C is not a part of any cluster.

So in that case would you suggest/is it possible to Install DB on Host A and CI+ASCS on HOST B(without ERS being installed anywhere)?

What if we do not install ERS, will it still be possible to build a system without ERS?

and even if we do install it,do we need to install ERS with CI(like you mentioned above),or ideally ASCS and ERS should be on different clustered nodes,like for example, Node A(DB node+ERS) and B (CI+ASCS node).?

The file system details provided by you has once again eliminated my certain doubts,and is very helpful.Thank you

Thanks a lot once again for your help and guidance.

Regards,

Bharat Pasricha'

former_member188883
Active Contributor
0 Kudos

Hi Bharat,

What if we do not install ERS, will it still be possible to build a system without ERS?

In absence of ERS there is a possibility of data loss during failover. ERS basically will hold the enqueue table entries. During failover these entries gets moved and updates are not lost during the process.

Hope this helps.

Regards,

Deepak Kori