HA SAP on Oracle

[Former Member]

Hello . 
I have a question related to HA configuration on Oracle database, may be some one tested scenario like this :

2 cluster with 2 nodes for 1 SID SAP installation.

On first 2 nodes will run only Oracle DB package and NFS shares (/sapmnt/SID)  second node for DA scenario if first node will fails.

On second cluster 2 nodes will be ASCS and ERS (may be PAS also on one of them). Application servers will be outside of cluster

on separate hosts.

Question , if database fails on 1 node , and cluster will move database on second node , what are the best solution -->

1) all applications serevers will wait with recconect status until databse will be available

2) all application servers will be restarted also (may be ASCS also, but this package not related to DB ).

Thanks for any comments.

[Former Member · Accepted solution]

Hello,

We have this kind of scenario implemented :

OS : IBM AIX 6.1

Database : Oracle, MaxDB

Node 1 : ASCS, SCS, DVEBMGS running

Node 2 : ERS and Dialog instance running.

During the failover ASCS & SCS(message server) resources are moved from Node1 to Node2 and Dialog instance is already connecting to message server so no need for re-start or re-connection.

This is tried and tested solution and it's working fine for us.

Regards,

Nitesh

[Former Member · Accepted solution]

Hello Nitesh , can you clarify on which node you have database ? I'm asking about database
rellocation not ASCS .
Regards , Sergo.

[Former Member · Accepted solution]

Hello Sergo,

DB is on Node1 which also is automatic moved from Node1 to Node2 during failover.

Regards,

Nitesh

[Former Member · Accepted solution]

Thank you Nitesh , can you please check how this parameters configured on application servers >

rsdb/reco_trials

rsdb/reco_sleep_time

rsdb/reco_add_error_codes.

And you have already some problems with DB (DB server) during high load and see what after database moved to second node all works fine (all update tasks,batch jobs and so on) or you only tested during premilinary tests then nothing happened in system ?

If you see what it works during high load you faced some problems or all was fine? Thank you, regards Sergo.

[ACE-SAP]

Hello

It would be good to know the kind of Oracle HA/cluster option => DataGuard, Failsafe, RAC ?

As far as tnsnames configuration file provides  connection information for all DB Nodes (multiple ADDRESS defined), SAP Application servers will automatically try to reconnect  (DB reconnect behavior of SAP system) and will try other DB nodes (native Oracle client behavior).

You do not need to restart all application servers.

SID.WORLD =

(DESCRIPTION =

    (ADDRESS_LIST =

(ADDRESS=(COMMUNITY=SAP.WORLD)(PROTOCOL=TCP)(HOST=Node1)(PORT=1527))

(ADDRESS=(COMMUNITY=SAP.WORLD)(PROTOCOL=TCP)(HOST=Node2)(PORT=1527))

    )

(CONNECT_DATA = (SID = SID)  (GLOBAL_NAME = SID.WORLD)

)

)

Best regards

98051  - Database Reconnect: Architecture and function

562403  - FAQ: Oracle Net

[Former Member]

Thank you Yves. For HA will be used HP Service Guard solution , with virtual name and IP rellocation.
On second node DB will be started with same IP and virtual hostname.
Which problems we can faced in this scenario? I'm not sure if during high activity user load , for some reason databse will be down on first node and will try to start on another , what will be happend with update requests, with batch jobs (also /sapmnt/SID/global will be relocated and will be no available some time), with all user requests ? I read all notes regarding reconnect , i dont need references to notes ... May be some one already tested scenario like this ? Thank you in any case.  My opinion is to restart all dialog servers after database restart , but if rellocation will work correctly also it is one of the way ....