ADP on new system

[JMALMAGRO]

Hello!

I'm trying to configure the ADP on a newly installed FAS2552 system.

The internal disks are 20 x 900 SAS+ 4 x 200 SSD

I did this procedure to change it from 7-mode to C-DOT 8.3RC2 https://kb.netapp.com/support/index?page=content&id=1013517

I read that the partitioning procedure is simply destroy the aggr0, unassign all disks, and select option 4 on Special Boot Menu.

But previously to zero disks, I get this message:

Mar 17 20:14:56 [localhost:raid.autoPart.disabled:notice]: Auto-partitioning is disabled on this system: the system needs a minimum of 4 usable internal hard disks.

I don't understand what this message means, nor if I'm  doing something wrong, because I didn't find any information for this ADP procedure.

Can you help me ????

Thanks in advance!

[f_defrocourt]

Hi JMALMAGRO,

The procedure to make FAS25XX in ADP mode :

http://kb.netapp.com/support/index?page=content&id=1015004&actp=SUBSCRIPTION&access=s

Regards,

François

[YIshikawa]

Remove 4 SSDs from slots and try initialization again.

[JMALMAGRO]

Hello! Thanks for your answers

The scenario right now is:

- Controller A took many SAS disks, partition all disks and created aggr0.

- Controller B took 3 SATA disks (from shelf 1) and didn't partition this disks. 

Question is: ADP is supported on external shelves? I read the limit for 2 external shelves plus internal "shelf". So this may be supported, except if SATA disks are not suported by ADP (I don't know if this is true)

The documentation at low level is very poor or inexistent ( I think it's normal, CDOT 8.3 is very new) and I cannot find specifically the answers to this concrete questions.

I'm going to create one SVM for NFS (Virtual Machines) and another SVM for CIFS. So, with this scenario, the idea I came with I don't know if applies: 1 node with all SAS, and 1 node with all SATA. As I can partition SAS disks, may be a good idea to give both controllers SAS and SATA disks, SAS for root and data, and SATA only for data.

But if SATA disks can be partitioned, I'll continue with the first idea (SAS for one controller, SATA for another, and SSD shared with a Storage pool)

What do you think?

Thanks!

PS: Sorry for my bad english

[YIshikawa]

As far as I have heard, Root-data partitioning only supports entry-level FAS or AFF configuration. External shelves are not supported, and disk assignment is not configurable. Disk assignment in ADP will be decided based on number of internal disks. Half of internal disks are assigned to controller A, and another half is assigned to controller B. Internal SATA disks are also supported in ADP.

[cedric_renauld]

Hi, 

Like the previous answer ... The ADP is closed in the contrloler,but the ADp is just for theRoot partition 

So you can have the root part for the tow controlersin the controler disk SLICE ROOT, and the first controler dan have the big SLICE ooff all disk inthe controler and the second controler have the shelves for dataexen if it's root part is Slice on the controler disk 

[DHESSARROW]

I know this is an old thread but I wanted to clear up some confusion for anyone stumbling on it.

ADP does require internal disk to function however the ADP aggragate can be extended to extra shelves, (last I checked you had to add the disk via CLI)

You also DO have the option of changing ownership of the partitions, most of my notes are from 6-8 months ago so things may have changed a bid and there may be good docs out there now.

examples below

"Set –priv advanced" is requires to make changes.

Cluster::*> storage disk show -disk 1.0.2* -partition-ownership
                                 Root                      Data        Container     Container
Disk     Aggregate Root Owner    Owner ID    Data Owner    Owner ID    Owner         Owner ID
-------- --------- ------------- ----------- ------------- ----------- ------------- -----------
1.0.2    -         Node-02       53691319   Node-02       53691319   Node-02       53691319
1.0.20   -         Node-02       53691319   Node-02       53691319   Node-02       53691319
1.0.21   -         Node-01       53691318   Node-01       53691318   Node-01       53691318
1.0.22   -         Node-02       53691319   Node-02       53691319   Node-02       53691319
1.0.23   -         Node-01       53691318   Node-01       53691318   Node-01       53691318
5 entries were displayed.

Cluster::*> storage disk assign -disk 1.0.20 -data  -owner Node-01

Error: command failed: Failed to assign disks. Reason: Disk 1.0.20 is already owned.

Cluster::*> storage disk assign -disk 1.0.20 -data  -owner Node-01 -f

Cluster::*> storage disk assign -disk 1.0.22 -data  -owner Node-01 -f

Cluster::*> storage disk show -disk 1.0.2* -partition-ownership
                                 Root                      Data        Container     Container
Disk     Aggregate Root Owner    Owner ID    Data Owner    Owner ID    Owner         Owner ID
-------- --------- ------------- ----------- ------------- ----------- ------------- -----------
1.0.2    -         Node-02       53691319   Node-02       53691319   Node-02       53691319
1.0.20   -         Node-02       53691319   Node-01       53691318   Node-02       53691319
1.0.21   -         Node-01       53691318   Node-01       53691318   Node-01       53691318
1.0.22   -         Node-02       53691319   Node-01       53691318   Node-02       53691319
1.0.23   -         Node-01       53691318   Node-01       53691318   Node-01       53691318
5 entries were displayed.

I also found that if you attempt to add a non partitioned disk to a partitioned aggregate in the cli it will convert it to a partitioned disk on the fly.

[rajesharumugam]

I have Newly installed 2552 controllers with ADP . i am getting only 13.75 TB but in the dashboard its showing 16.82 . where is the rest of the sapace...

please find the attached image

details: 

2552 controllers (dual)

ADP configured with one large aggrigate ( allocated all the disk to one controller)

only root avalable on other controller.

Thanks in advance

[SeanHatfield]

The "Disks" box is showing raw disk numbers.  Within the aggregate 2 data slices are used for RAID parity, and 10% of the remainder is reserved for WAFL.   

[rajesharumugam]

Hi Sean,

   Thanks for the clarification but there is no major space saving comparing to traditional setup and ADP setup i am getting only 200 GB more...

[SeanHatfield]

To get that close you have to run without spares, and you have fewer spindles providing iops to the data aggregate.