ORACLE GRID 11gr2 (RAC) on HPVM Guests

[RafS]

Hi all,
I installed ORACLE GRID 11gr2 (RAC) on 2 HPVM guests v4.2.
The VMs reside on 2 different HPVM Hosts on 2 c7000.

When I try to start the instance, it takes about 15 minutes to start up.

Using the command :#netstat -s -p ip, I get too many "fragments dropped after timeout".

Have you ever installed ORACLE GRID 11gr2 (RAC) on HPVM Guests ?
Have you any ideas to solve this problem?

Thanks a lot,
Raffaele

[Duncan Edmonstone]

Ummm you do know that RAC isn't officially supported in a HPVM don't you?

http://www.oracle.com/technetwork/database/virtualizationmatrix-172995.html


It should however "work" - Where are you seeing the "fragments dropped after timeout"? on the guest or on the host?

HTH

Duncan

[RafS]

Hi,
I know that RAC isn't supported, but I think it should work.

I see the fragment dropped on the VM.

Thanks a lot,
Raffaele

[Duncan Edmonstone]

So do you see anything on the VM host?

HTH

Duncan

[RafS]

On the VM Host the fragments dropped after timeout are 0.

Thanks a lot,
Raffaele

[Duncan Edmonstone · Accepted solution]

What AVIO driver version do you have in the host and guest? Are you on the latest and greatest?

https://h20392.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=AVIO

HTH

Duncan

[RafS]

Hi,
I have:
HostAVIOStor B.11.31.1003.01
HostAvioLan B.11.31.1003.01
GuestAVIOStor B.11.31.1003
GuestAvioLan B.11.31.1003

I'll upgrade AVIO software and I'll give you any news.

Thanks a lot,
Raffaele

[RafS]

Hi all,
now I have:

GuestAVIOStor B.11.31.1103
GuestAvioLan B.11.31.1103
HostAVIOStor B.11.31.1103
HostAvioLan B.11.31.1103.01

but the problem persists.

Thanks again,
Raffaele

[Duncan Edmonstone]

OK, so have you reviewed all the known networking issues in section 7.2 of the HPVM4.2 release notes:

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c02471305/c02471305.pdf

HTH

Duncan

[RafS]

Hi,
I upgraded the VM from 4.2 to 4.3, but I didn't solve the problem.

Thanks again,
Raffaele

[grahamswilson]

...a couple of things...

1. Even though it isn't really "supported", I know it does work, as I have a two node 11gr2 RAC on 2 HPVM guests running v4.2...so I can confirm it can be done!

2. Not sure if it's relevant to your issue, but how is the "private" interconnect for the RAC heartbeat/HA configured? Did you use a local vswitch or use the public interface? I have had issues in the past with UDP packet fragmentation over this private link, and on physical servers we had found one way round this was to enable jumbo frames on the switches. In my HPVM RAC enviroment, I used a local vswitch and never had to make any changes...not yet anyway!

[RafS]

Hi,
we have 2 VMs on 2 different Hosts, so we can't use local switch.
We enabled jumbo frame on the physical switch on the Host's NIC and on the VMGuest's NIC


thanks,
Raffaele

[Steven E. Protter]

Shalom Raffaele,

The reason Oracle support is important to consider is that after you resolve the OS issue you are facing, you may at some point need Oracle support for the application.

As soon as you reveal the unsupported configuration, Oracle is very likely to tell you the environment is not supported and they are sorry but they won't help you.

Oracle RAC does require a private heartbeat network. This can be done through a virtual interface but may not be the most stable way to configure it.

SEP

[blam]

Hi,

I'm not running RAC but just single node mutiple homes Oracle database server with over 12 Prod databases currently on physical Itanium (HPUX 1131 Sep 2010) and planning to move that to HPVM very soon.

I'm a bit unsure what to expect after the migration in terms of performance

Any comments/ advice on that?

FYI on the HPVM I'll be running the latest HPUX 1131 Sep2011

Thanks for any advice/comment

Brian