JDE 9.0 and RAC

Does anyone have experience setting up 9.0 with an Oracle RAC database? Any RAC experience is helpful, RAC with 11 G r2 is most helpful. Here is our problem:

We have have two Exadata servers (Sun Sunfire servers). The Exadatas have multiple compute nodes hosting Oracle 11 G r2 databases. We use RAC to keep the data in synch between the two nodes. Rac requires two different TNS host names for each server.

If both servers are up and listening, we can swing JDE back and forth between the nodes. If one or the other node is down, then either our login time is delayed (one minute of delay) in our best case scenario, or we can't log in at all.

The clients are weblogic servers on Linux. The batch servers are windows 2008. We also have this issue with linux based batch servers. (we are trying out both windows and linux for the batch server layer of our cloud of servers)

On the windows/sql world, we do not have this issue. When we swing from node to node, the MS cluster names and the SQL cluster names remain the same, just the server hosting them changes.

We have an open ticket with Oracle on this, but they are stumped. We would assume that there must be some other customers out there that are using 9.0 with RAC. We don't think this is an Exadata issue, we think that this is a Rac issue.

Any thoughts?

below is the latest TNS that oracle had us try out:

pxprod.na.praxair.com =
(DESCRIPTION_LIST=
(LOAD_BALANCE=OFF)(FAILOVER=on)
(DESCRIPTION =
(LOAD_BALANCE=on)(TRANSPORT_CONNECT_TIMEOUT=5)(CONNECT_TIMEOUT=10)(RETRY_COUNT=0)
(ADDRESS = (PROTOCOL = TCP)(HOST = jdedb1-scan)(PORT = 1521))
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = jdedg.na.praxair.com)
)
)
(DESCRIPTION =
(LOAD_BALANCE=on)(TRANSPORT_CONNECT_TIMEOUT=5)(CONNECT_TIMEOUT=10)(RETRY_COUNT=0)
(ADDRESS = (PROTOCOL = TCP)(HOST = jdedb2-scan)(PORT = 1521))
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = jdedg.na.praxair.com)
)
)
)

pxdev.na.praxair.com =
(DESCRIPTION_LIST=
(LOAD_BALANCE=OFF)(FAILOVER=on)
(DESCRIPTION =
(LOAD_BALANCE=on)(TRANSPORT_CONNECT_TIMEOUT=5)(CONNECT_TIMEOUT=10)(RETRY_COUNT=0)
(ADDRESS = (PROTOCOL = TCP)(HOST = jdedb1-scan)(PORT = 1521))
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = pxdevdg.na.praxair.com)
)
)
(DESCRIPTION =
(LOAD_BALANCE=on)(TRANSPORT_CONNECT_TIMEOUT=5)(CONNECT_TIMEOUT=10)(RETRY_COUNT=0)
(ADDRESS = (PROTOCOL = TCP)(HOST = jdedb2-scan)(PORT = 1521))
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = pxdevdg.na.praxair.com)
)
)
)

- Gregg

Greg,

I have several clients on RAC and full HA stacks. I just completed RAC - 11g R2 on Windows 2008 - R2 - 64 bit on VMWare in Dec so you should have no issues on your system.

On of our present projects is RAC - 11g R2 on Linux. Actually implementing this at the big Wine Company just down the road from you.

Please look at Doc 646085.1 - this is the closest thing you may find.

You also need to be on a minimum tools of 8.98.3.1 for full TAF (Transparent Application Failover).

On SQL Server you would have a long delay as the nodes transfer information and the system would be "down" until the 2nd node started up.

On Oracle RAC there is NO delay. Nothing happens and things just keep running. Even on early 10g releases which does not support full TAF very little happens and certainly no one minute delay.

Colin

Colin

We are on tools release 8.98.41

Are you able to send over a tnsora to my email address for comparision? Any other hints or suggestions?

I forwarded that Oracle doc over to our Oracle DBA, thanks.

- Gregg

And Colin, as the Performance Lead consultant on said project, you can't get us switched over to RAC soon enough!

-Ethan

Re: JDE 9.0 and RAC and Dataguard

Let me expand my inquiry - is anyone using 9.0, RAC and Dataguard? Flipping between compute nodes on the exadatas isn't the issue. The issue is when we use Dataguard to flip from the Exadata in the data center over to the Exadata in the DR data center. Specifically, when we swing over and one of the Exadatas goes offline.

Any dataguard insights?

BTW Colin, the DBA got back to me, our Max_commit_propogation value for 11G is the default - 0

MAX_COMMIT_PROPAGATION_DELAYProperty Description
Parameter type Integer
Default value 0

Thanks!

- Gregg

[ QUOTE ]
Actually implementing this at the big Wine Company just down the road from you.



[/ QUOTE ]

Does the Big Wine Company hand out free samples? I'm going to need some after this project goes live in March.

Sorry, I don't have an answer, but the LOAD_BALANCE setting in your TNS is once OFF and once ON, almost on the same line - I just find this suspicious...

Greg,

Is this a client tnsnames.ora file? It may not be related but this looks like a pre-11g R2 file. Oracle RAC starting with 11g R2 is way different than before. Only the SCAN (Single Client Access Name) needs to be entered - not all the individual hosts.

Oracle 11g R2

E1PROD =
(DESCRIPTION =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = JDEORA-cluster.customer.com)(PORT = 1521))
)
(CONNECT_DATA =
(SERVICE_NAME = E1PROD)
)
)


Oracle 10g R2

E1PROD=
(DESCRIPTION =
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = server1.customer.com)(PORT = 1521))
(ADDRESS = (PROTOCOL = TCP)(HOST = server2.customer.com)(PORT = 1521))
(LOAD_BALANCE = yes)
)
(CONNECT_DATA =
(SERVER = DEDICATED)
(SERVICE_NAME = E1PROD)
)
)

I may need to do some more investigation but we've always done 2 node RAC's. In the tnsnames.ora you have one node set to LOAD_BALANCE=OFF but FAILOVER=ON. Is it the desire to use RAC only for Failover and not for load balancing (ie using it at DataGuard?).

If that's the case then I can honestly say that I'm not sure how well that will work or if anyone has tried it.

Colin

Re: JDE 9.0 and RAC and Dataguard

Ahhhh.........see my last post before I saw your DataGuard usage.

Yeah......... so you want to fail-over using DataGuard between 2 Data Centers that may be thousands of miles apart and you expect JDE not to notice?

RAC is certified and works on JDE and supports TAF.

Data Guard supports TAF but does it support it with JDE? Might be time to pick up the phone. My gut reaction is that if you decide to switch the database to another data center you'll be restarting services.

Hmmmm...........Data Guard with Exadata? You're not expecting it to go down are you?

Colin

Ethan,

my hope is that RAC wil improve performance but it won't. Just a bigger highway in this case that can hold more cars. The speed limit hasn't changed. RAC in your case will only add capacity not speed.

For performance we can tune many other things.

Colin