11gR2 RAC Architecture Picture

From my just finished RAC course – it was an LVC without whiteboard, which forced me to do the sketches on the computer. You might find it useful. It literally takes the whole day (at least) to explain it, though. And yes: It is oversimplified and some things are missing (I paint all the time in that course) 🙂

Just for me to memorize:

ohasd starts the agents

orarootagent -> crsd, ctssd …
cssdagent -> cssd
oraagent -> evmd, asm …

crsd starts the agents

orarootagent -> node vip, SCAN vip …
oraagent -> Local Listener, SCAN Listener, Diskgroups, Database, Services …

Addendum: The corporation kindly offered to translate my above amateurish sketch into a professional graphic which I gladly accepted. See the brushed up sketch below:

I like it – just seems to me that due to corporate identity requirements, we are a little short in colours 🙂

11g New Features, High Availability, RAC

Dieser Eintrag wurde erstellt am November 18, 2011, 18:56 und wurde abgelegt unter TOI. Du kannst die Antworten auf diesen Beitrag über RSS 2.0 verfolgen. Du kannst eine Antwort schreiben oder einen Trackback von deiner eigenen Seite schicken.

#1 von exadatas am November 18, 2011 - 22:40

good one!!!
#2 von exadatas am November 18, 2011 - 22:43

I see Server on Host A .. is there anything missing in Host B?
#3 von Uwe Hesse am November 18, 2011 - 22:51

That is a server process, attached to the instance orcl1 and connected to the user process on the client. There are typically many of them on both nodes – I just use that one to explain failover etc.
#4 von Ora600Tom am November 19, 2011 - 05:50

Nice one, thank you. You have placed the voting disk and OCR along with the other files – so used EXTERNAL redundancy with one copy of both OCR and voting disk. Is this a recommended way? I understand, any disk failure will be covered by RAID, so we are safe. But, what about any file corruption or logical failure? Will it bring down the whole cluster down? So, instead of going for EXTERNAL, is it worth considering NORMAL?

But, I do agree, EXTERNAL redundancy is one way.

Thanks
Thomas Saviour
#5 von Uwe Hesse am November 19, 2011 - 10:32

Thank you, Thomas, for your comment! The diskgroup DATA on the picture may have any redundancy (external, normal, high), which would then determine the protection by ASM of Voting File and OCR. In our course, it has redundancy normal, so there are three voting files and the OCRs stripes are mirrored once. That is by the way also the way we do it on Exadata (at least redundancy normal for the diskgroup that contains Voting Files and OCR). If your storage is already delivering redundancy – go with ASM external redundancy. That is true for any diskgroup.
#6 von Ora600Tom am November 19, 2011 - 15:05

Thanks a lot for clarifying.

Thanks
Thomas Saviour
#7 von lotfibouziani am Januar 5, 2012 - 17:37

Hi Uwe,
Thanks for all what you share.
I need to understand the network acces (using SCAN, GNS and DNS) in RAC 11gR2. Do you have any schema to simplify this concepts please ?
My email is : lotfi.bouziani@gmail.com
Best regards.
#8 von Numz Sittipol am April 30, 2012 - 11:18

Hi Uwe,
I have some question about OLR. What is the exact information that OLR keep?
If I store my OCR in ASM diskgroup, when ASM instance containing my OCR cannot start, how clusterware identify my OCR location using OLR?

Thanks
Sittipol.
#9 von Uwe Hesse am April 30, 2012 - 11:56

Sittipol, the OLR stores the Local Resources. You can see them with crsctl stat res -t , which differentiates between Local Resources and Cluster Resources. Another way to see more detailed the contents of the OLR is ocrdump -local

The OCR cannot get accessed unless the ASM instance on the particular node has mounted the ASM diskgroup on which the OCR resides.

But the voting disk can get accessed before, because the OLR points to the ASM disks that hold the (not striped) voting disks. In other words: We can access the voting disk already before the ASM diskgroups have been mounted.
#10 von Numz Sittipol am April 30, 2012 - 12:37

Many thanks for
#11 von Numz Sittipol am April 30, 2012 - 12:39

your quick response. Your answer is very clear.
#12 von Arun am Mai 8, 2012 - 11:45

ur just superb
#13 von SC Chen am November 27, 2013 - 17:01

SAN storage does not connect using NIC network card, unless it is NFS, iSCSI. Moreover, this is a point of failure if the network cable failed. There should be 2 NIC from each servers
#14 von ls am Juni 14, 2014 - 06:01

nice
#15 von Sachin am Juli 14, 2015 - 13:33

Hi , I have one doubt if olr is used for voting disk or ocr location before the asm startup… Than what is the use of grid plug and play profile.. Gnpd also contain the disk info and used for the same… Please make me clear… Thank you