Yesterday I gave an online seminar for the RAC SIG (http://www.oracleracsig.org), based around the “What’s So Special about the Oracle Database Appliance” presentation both Lonneke and I have presented.
I was very conscious that I was talking to an esteemed audience and added some extra slides with information that I thought that experienced RAC administrators might appreciate, like DR. Here are the slides (current at the time I’m writing this):Watch movie online Logan (2017)
What was encouraging was that there were plenty of questions – around 20 minutes or so I think. Most I answered, though some other experienced attendees also threw in some good responses. I thought it might be worth summarising them here, partly for anyone who missed the webinar (though it should be available for replay soon from the RAC SIG website), and partly so that I can improve the slides for the future (I know they’ve got a bit unwieldy now!) .
So here goes; I copied this from the WebEx chat window and will try to answer as I did at the time (given that I don’t have a recording of myself!).
Indira K: I would like to note down the doc URL in the previous slide 18 please?
A: This is the Oracle’s handy ODA set-up poster which comes in the box. The PDF version for the ODA X4-2 is at: http://docs.oracle.com/cd/E22693_01/doc.21/e41668.pdf and you can find the overall ODA docs at: http://docs.oracle.com/cd/E22693_01/index.htm
J Antonio M: Differences between EXADATA/EXALOGIC with ODA… Any additional benefits?
A: I’m not sure I actually answered this question! I think I illustrated the key technical differences in slide 46, but perhaps could spend some time speaking about non-functional differences (i.e. how you “live” with these machines) another time. The benefits of ODA vs Exadata/Exalogic, and which suits you better, will depend on your own particular requirements.
General Q/A: There was a bit of discussion about how many nodes/servers an Exadata 1/8 had – the slides were correct and Govi provided this link: http://www.oracle.com/technetwork/database/exadata/exadata-dbmachine-x4-2-ds-2076448.pdf (i.e. 2 database servers plus 3 storage servers).
Ravi A: Would ODA be recommended for OTLP or OLAP env?
A: I said I didn’t really know – presumably mostly OLTP but I had heard of a customer connecting their ODA to a ZFS appliance to get HCC features for a data warehouse. Ravi Sharma, from Oracle, also replied: “ODA works for both workloads. Please also see the white paper: http://www.oracle.com/technetwork/server-storage/engineered-systems/database-appliance/documentation/oda-eval-comparing-performance-1895230.pdf”
Sravan: Can we applying rolling patches using OAKCLI – one node at a time? and
Shiju J: Can we patch OS, GRID and Database at 3 different windows one at a time?
A: I know the bundle patches for infra need reboots but didn’t know if both nodes had to be down at the time. I don’t know if you can do patches for the different layers concurrently but I don’t think it’s a good idea even if Oracle doesn’t forbid it – I think you want to follow the most common path for other users (as this is therefore the best tested one) which will be to do them sequentially.
Ravi Sharma (Oracle) added “No you should upgrade OS, GI, and DB sequentially.”
Ranjit: Is oracle planning to patch one node at a time in near future?
Ravi Sharma (Oracle) replied: “We currently do patch one node at a time, but I think your question is related to rolling patches for everything. The answer is no at the moment.”
Shiju J: Can we do OS on one weekend and then GI on a different weekend ..to reduce the down time?
Ravi Sharma (Oracle) replied: “Yes, you can do the INFRA, GI, and DB patches at different times (for example separate weekends).”
Yuva: How long did it take to see the first disk failure? How long to get the replacement disks?
A: That was the first one last week for the database case study (2 ODAs each with dual arrays, so out of 88 HDDs and 16 SSDs). I’m not involved with their operations so don’t know how long it took for Oracle to replace it in practice.
Vishal: Does this support jumbo frame within private network for RAC?
A: Yes it does – and I checked the MTU specifically and on the interconnect it is set to 9000 bytes by default. You can also configure jumbo frames on the other networks, though not automatically from a CLI (last time I checked). For my network investigations see http://www.veriton.co.uk/roller/fmw/entry/virtualised_oda_poc_3_networking
Vishal added: That’s great if MTU gets up to 9000
Paul B: If I want to use RAC with 4 cores per node, do I license 4 or 8 cores across the board in total?
A: The licensing rules apply as normal, i.e. if you have full RAC then you have to fully license the cores used on both nodes. If you are using RAC One Node you may be able to license just one node under the 10 day fail-over rule.
Deepak M: What is the group name on LinkedIn?
A: It’s called the “Oracle Database Appliance User-Group” and you can find it at https://www.linkedin.com/groups?home=&gid=4348075
Keyur: Can we install Oracle EM on this… 2 VM for DB and 2 VM for EM?
Keyur: Is there a ready to use OEM VM template for ODA?
A: That was a really good question that I should probably cover in the slides. Recently Oracle released a special template for ODA with EM in it (currently it is EM 12c R3 but I assume it will be updated to R4 soon). There’s a white paper to go with it on OTN and a document in MOS.
Ravi Sharma (Oracle) added: “Yes, we recently released a white paper and a template for OEM on ODA. Please see http://www.oracle.com/technetwork/database/database-
Yuva: How can we expand the storage space on X3 ver1 to our existing SAN?
A: If it’s a Fibre Channel SAN I don’t think there’s any way that can be done. If it’s iSCSI it may be possible but check MOS as there’s a note on there about expanding ODA storage (e.g. to ZFS and NFS arrays) and that may say.
Siva: ODA – does it have limitation of only 2 nodes RAC?
A: Yes, there are only two nodes in the ODA, and on the virtualized platform only two database VMs. You cannot cluster multiple ODAs together to form some sort of giant cluster.
Rajat B: In case of full system crash (one node) including OS… what would be the major recovery steps… not database alone?
A: I wasn’t entirely sure about that – backup of node using OS utilities perhaps. For VMs on the (ACFS provided) shared storage there are other options, but this doesn’t apply to the ODA Base VM as that is on local storage.
Ranjit: There were some error in patching to 2.7 where we need to apply GI patch introduced by INFRA patchWatch Full Movie Online Streaming Online and Download
Ravi Sharma (Oracle) replied: “Yes, generally it is recommended to patch INFRA and GI together (esp if there is a dependency for a particular patch).”
Vishal: Can we dynamically increase CPU for server what’s the limit to that?
Yes – it depends on the ODA mode.
If it is Bare Metal you have a licence key from MOS which sets the cores available. As it’s at the BIOS level it won’t be dynamic, but it may be possible to do it in a rolling fashion (you’d need to check).
If it’s ODA Virtualized Platform then you set up the cores (and memory) when configuring the ODA base and can change that later. Again I don’t know if that can be done rolling or not.
Vishal: Anything to control buffer size at hypervisor?
A: I guessed this was related to memory usage by Dom0. You need to leave some memory aside on Dom0 (8GB I think is recommended).
So, thanks again to everyone who joined the webcast – if you’d like to chat more about ODA please get in touch and perhaps a few of us can meet up at Oracle OpenWorld!