Reconfiguration to run central instance on Cluster.

[Former Member]

Hi All,

We have Netweaver 7.0 installed on Sun Solaris and Oracle as database.We have a cluster configured but our central instance (JC01) is not failing over to node2 on failure of node1 while database and central services (SCS) has no issue on running on either node.

we performed a check and found that at the time of installation Central Services Instance (SCS) was installed on virtual node (SAPPLPNW) while Central Instance (JC01) and Dialog Instance (J01) were installed on SAPPLPR1 and SAPPLPR2 respectively.Also below  profile parameter justified the same configuration.Below are the profiles we have :

START_SCS00_sapplpnw – Start Profile for SCS

START_JC01_sapplpr1 - Start Profile for JC01

START_J01_sapplpr2 - Start Profile for J01

PPR_SCS00_sapplpnw – Instance Profile for SCS

PPR_JC01_sapplpr1 – Instance Profile for JC01

PPR_J01_sapplpr2 – Instance Profile for J01

Start profile is using SAPLOCALHOST = sapplpr1 and Also Start profiles is calling PPR_JC01_sapplpr1 instance profile (  _PF = $(DIR_PROFILE)/PPR_JC01_sapplpr1) where again host name is SAPPLPR1 ( SAPLOCALHOST = sapplpr1).

Whereas this is not the case with SCS , in both Start and Instance profiles SAPLOCALHOST = sapplpnw so SCS has dependency on virtual host and can be started on either node depending upon the location of virtual host (i.e. on which host this service is running).

Now we want to do the reconfiguration to make sure that central instance is able to run on cluster and can failover to node2 like cetral services (SCS) and Database. Please suggest what options we have to do this reconfiguration.

Thanks and Regards

Sachin Gupta

[former_member189725]

I suggest you change the hostname from sapplpr1 to sapplpnw manully for JC01. You need to rename the profile names (instance and start) . Then edit both these profiles and change any reference of hostnames. Then open the configtool. Switch to offline configurator  mode . Change the cluster data for the instance. Also check if there is any hostname reference in the configtool by searching with old hostname and replacing with the new one.

This has a disadvantage as the instance ID (ID_XXXXX) will remain the same even though it should be different which is generated using the box number ( defined using hostname,SID and instance name).

But this is the fastest way you can include the system in the cluster group for failover to happen.

Otherwise you can re-install using the system copy method which will take a lot of downtime.

Also note the places where you have made the changes.

Regards

Ratnajit

[Former Member]

This approach of simply renaming profiles and params won't work for JAVA systems. A new install either fresh or using system copy would be needed. JAVA has quite a few system files at OS which have reference to the hostname and it is not like ABAP where just changing profile names and parameters will do the job.

Coming to the main question so called Central instance need not be a part of cluster as such. We only cluster SPOF not things which can be made redundant. CI ijn JAVA can be reduntant except for probably SDM/JSPM and couple of things ( idon't recal exaclty)  but deploying additional secondary application servers. SAP doesn't really recommend CI HA.

Personally I have implemented CI over HA but when I queried SAP about that there was no clear cut response as to whether they are gonna support it or now till the time I was in that particular implementation But from the vibes I got CI on HA isn't something SAP is too happy about.

Regards.

Ruchit Khushu

[former_member215961]

Hi Sachin,

As far as I know Hight Availavility Arquitecture it is described that the Central instance and Dialog instance has to be installed on local disks . So as you have described you are to the right way.

The (A)SCS instance is clustered makes balancing to one node or the other. But Central instance or dialog instance cannot.

I don't work much with Solaris Arquitecture but I've implemented a similar scenario with Red Hat.

There is any approaching to make that CI and DI failover over the cluster , for this reason are installed on local disks. Let me explain it with an example :

( Before Failover)

Node A has resources: A(SCS), ERS(node A) and CI running

Node B has resources: Database Instance, ERS(node B) and DI running.

On a Failover situacion for node A: (A)SCS resource takes offline to node A as well ERS( node A ) and CI . Then A(SCS) brings online to node B.

( After Failover)

Node B has resources: A(SCS) , Database Instance, ERS(node B) and DI.

Dialog instance might need to reconnect to Message server maybe there's a production downtime no longer than 30 segons.

Another matter was if you would like to brings online DI at Failover occurrence, Then your  Cluster software can be usefull to implement it. Search a REDBOOK or a HA Guide for your cluster software vendor, I'm sure that it will be usefull.

hope this helps you.

Cheers.