Re: Migrating from 7 mode to cDOT - help with cluster design

irakli_natsvlishvili · ‎2016-07-14

Hello everybody,

After running with 7-mode for >8 years we have to move to clustered ONTAP and we would like to do it with minimum disruption and need to re-engineer our tools/scripts.

In our current setup we have FAS8020 with dual controllers and 3 24disk shelves per controller, with identical disks. We have just one aggregate per controller, where we have root volume and data volumes (less than 10) as well. Most volumes are used for NFS3, while couple of them - with CIFS. We have 2 1G and 4 10G port per controller which we use following way: one 1G for CIFS (VLAN 10), on1 1G - for administration (VLAN 30), two 10G interfaces for NFS combined in МША with a single VLAN (VLAN 30), two remaining interfaces companed into aggregated channel (VIF) with VLANs enabled for separate VLAN access.

So my question is - how much it is possible to adapt similar design with cDOT? If my understanding is correct, cDOT requires separate root aggrage and we can host data SVMs/volumes at root aggregate. Is this correct? If not - how can I change the standard setup?

Second question - about network. While we plan to have switchless cluster, it would take 2 interfaces (1 - 1G and 1 - 10G) for cluster interconnect so interfaces available for project/administraton networks are reduced. Is this statement correct?

Finally question about SVMs and volumes. At our current setup we have the same volume accessible from different network/purposes. How can I setup the volumes/SVM the same way?

Finally - is there any document/note/report which would help me to understand differences between cDOT and 7-mode and make our transition painless?

Thanks.

aborzenkov · ‎2016-07-15

If my understanding is correct, cDOT requires separate root aggrage and we can host data SVMs/volumes at root aggregate. Is this correct?

Yes and no (unless you made a typo). cDOT requires separate root aggregate and it is highly discouraged to host SVM data on it. It is dedicated to root volume only.

While we plan to have switchless cluster, it would take 2 interfaces (1 - 1G and 1 - 10G)

Wrong. cDOT requires two dedicated 10G ports for cluster interconnect (switched or switchless does not matter). Everything else can in principle share the same remaining physical ports; it depends entirely on your network connectivity.

Finally question about SVMs and volumes. At our current setup we have the same volume accessible from different network/purposes. How can I setup the volumes/SVM the same way?

Well, if you want to avoid redesign - migrate each 7-Mode filer to own SVM keeping the same names and addresses. This will give you familiar management model where you have two independent servers with own resources. Clients won't notice anything.

Finally - is there any document/note/report which would help me to understand differences between cDOT and 7-mode and make our transition painless?

You could start with http://transition.netapp.com/. But judging by your question, I highly recommend to arrange for professional service, either from NetAp or from NetApp partner, to plan and implement such migration.

irakli_natsvlishvili · ‎2016-07-20

>cDOT requires separate root aggregate and it is highly discouraged to host SVM data on it. It is dedicated to root volume only.

On release notes of 8.3.2 says following

Starting with Data ONTAP 8.3.1, Data ONTAP enforces the rule that the root aggregate be dedicated to only the node's root volume. Data ONTAP prevents you from creating other volumes in the root aggregate.

So are you telling that it is possible to have SVM on root aggregate or not?

> Well, if you want to avoid redesign - migrate each 7-Mode filer to own SVM keeping the same names and addresses.
> This will give you familiar management model where you have two independent servers with own resources. Clients won't notice anything.

Not possible. I have 16 volumes on a an aggregate on one filer, another filer has 12 voilumes. Files on each of volumes are used for flexcloning via automated provisining via scripts. It means that it have to rewrite these scripts and continue modifying them when adding/removing new volumes.

As an example. Thre are two volumes - Vmware_NFS_1 and Vmware_NFS_2. Fist should be accessible from ESX servers from network of 10.10.10.0/24 via dedicated 10G port, and the second volume should be accessible via NFS from the same network and twh other networks 10.10.20.0/24, 10.10.30.0/24, plus via CIFS from network 192.168.10.0/24. last three networks should access the volume via second tunk port which has VLANs enabled as well.

Above setup at current 7-mode is done easy - one aggregate, two volumes on the same aggregate, two NFS exports - they define clients from which network have access rights, one CIFS share and two network ports used on filer - e0c for 10G dedicated NFS access via 10.10.10.0/24 and e0d as a trunk port with VLANs 20, 30 and 40 which respectivly are for 10.10.20.0/24, 10.10.30.0/24 and 192.168.10.0/24.

I want (need) above setup migrate to cDOT with minimum redisign. Is it possible?

aborzenkov · ‎2016-07-20

So are you telling that it is possible to have SVM on root aggregate or not?

It was (and I'm pretty sure still is) tecnically possible. That current Data ONTAP makes it more difficult to do by accident is good. Once again - doing it is not and never was recommended.

I want (need) above setup migrate to cDOT with minimum redisign. Is it possible?

Yes, of course. Do you have specific questions or concerns?