RAC Hardware and OS Considerations 1

[tekdudedude]

Hello,

Does RAC require all your servers to be the same hardware and\or OS version?


Thanks,

TekDude

 

[Thargy]

No, yes

http://download.oracle.com/docs/cd/B28359_01/rac.111/b28252/intro_tdprc.htm

refers.

I quote

Oracle Clusterware and Oracle RAC do not support heterogeneous platforms in the same cluster. For example, you cannot have one node in the cluster running Red Hat Linux and another node in the same cluster running Solaris UNIX. All nodes must run the same operating system, that is, they must be binary compatible. Oracle RAC does not support machines having different chip architectures in the same cluster. However, you can have machines of different speeds and sizes in the same cluster.

End quote.

Regards

T

Grinding away at things Oracular

 

[tekdudedude]

thargtheslayer,

Thanks for the response.

Just to clarify RAC nodes having to be on the same OS major version:

1. On a two node RAC, can one node be RH ES 4 and the other RH ES 5?

2. In MS terms on a two node RAC, can one node be Windows 2000 Server and the other Windows 2003 Server?


I think this is good to know before implementing RAC into production.

Thanks again,

TekDude

 

[Thargy]

tek,

are you in the uk?

Me, I'm sl;umming it at the keyboard chez moi.

Can we just chat about this? It's a bit involved, and would take too long to clarify via text.

Can you send me your email address in the form of tek at tekaddress dot com?

T

Grinding away at things Oracular

 

[Thargy]

tek,

failing that, go to a thread with SantaMufasa in it, and he can give you me email address.

he can read this thread to confirm that it's ok for him to give it to you.

Fair enough?

T

Grinding away at things Oracular

 

[tekdudedude]

thargtheslayer,

Thanks again for the response. No I am not in the UK. Sorry I missed your kind offer yesterday.

I was hoping to find in the Oracle docs a page that details acceptable OS RAC arrangements but I have not found it yet.

I have used RAC in Oracle 9i (Solaris). We are now considering for LINUX and Windows. I just want to clariy the options relating to scalability (as mentioned earlier).


Thanks,

TekDude

 

[Thargy]

cutting a long story short, I recommend that you make each node identical. Mixing even minor versions is asking for trouble, since it makes definitive fault finding almost impossible. If you fix something on one node, does that guarantee a fix on them all, if those other nodes are all slightly different?

I suggest identical nodes to KISS, as far as possible. Clusters seem to me anyway, tough enough to manage without any differences. Adding further complexity to the mix is just yucky, to use a highly technical term.

Regards

T

Grinding away at things Oracular

 

[tekdudedude]

thargtheslayer,

Thanks for the suggestion sir.

TekDude