RAC & Service Guard

[jpcast_real]

Hello Everyone,

We have an HP-UX server 11i with an MC/ServiceGurard Oracle Package. Now we are thinking to move to Oracle RAC and some questions come to my mind:

- Can I use for the RAC different nodes with different number of CPUs, memory etc, or all the HW must have the same characteristics.


- Our servers are placed 3 kilometers far away one from the other. How can a measure correctly the latency between both servers. Could this latency affect to the RAC??

- Our SG release is the 11.16 . Has anyone the figure about how much is the upgrade to SG extension for RAC.

-Is it possible to divide the load divided between both nodes of the RAC in case the power of this 2 nodes is not going to be the same.

Many thanks

[Oviwan]

Hey

If you have Oracle 10g use ASM instead of SG that's "free".

Regards

[jpcast_real]

Hello,

unfortunately we still can not move to Oracle 10g, so I wonder if it is extremely necessary to buy the HP extension or it can be manually programmed.

Regards

[melvyn burnard]

Firstly, "scripting" the SGeRAC is NOT an option, as it is not actually any scripts, but rather bits that now allow you to use the SLVM (Shared) rather than normal or exclusive mode LVM, and other bits.
You can certainly look at using nodes that are not the same cpacity, but this rather defeats the object of "sharing" the load across different active instances.

for pricing, take a look here:
http://h20293.www2.hp.com/portal/swdepot/displayProductsList.do?category=HA

Also take a read of the Support Matrixes and White Papers at:
http://docs.hp.com/en/ha#Serviceguard%20Extension%20for%20Real%20Application%20Cluster%20%28ServiceGuard%20OPS%20Edition%29

[Steven E. Protter]

Shalom,

Answers imbedded.

- Can I use for the RAC different nodes with different number of CPUs, memory etc, or all the HW must have the same characteristics.


Yes they can have different number of cpu's but they must be all the same type of cpu and the OS must be the same. A cluster can't contain a mix of say ia64 and pa-risc processors.


- Our servers are placed 3 kilometers far away one from the other. How can a measure correctly the latency between both servers. Could this latency affect to the RAC??

RAC might not like that distance. You might need HP's continental or metro cluster here. The connection between the servers has to be very past. Latency can be measured with network tools but should be tested with RAC.



- Our SG release is the 11.16 . Has anyone the figure about how much is the upgrade to SG extension for RAC.

I like 11.16. It's very, very stable and has most of the good security tools. I see no reason to move to 11.17. The service extension for rack is not an upgrade, its an add in that works with SG 11.16.

-Is it possible to divide the load divided between both nodes of the RAC in case the power of this 2 nodes is not going to be the same.

Yes, but the real problem is RAC requires shared storage. Where is the storage going to be, at node 1 or node 2. I'd immagine some latency issues running a cluster like that.

SEP