Adding disk shelf with root volume to other filer

[rajdeepsengupta]

Hi,

I have two questions.

1. I have one 3240 cluster with one header is having SAS disk(DS4232), and the other header with SATA disk(DS4243). Now I have to shift few FC(DS14MK4)  and SATA(DS4243) shelves from 3020 and 3040 headers. Now the idea is to add FC disk shelf to the 3240 header which is having SAS shelf, and the SATA shelf to the 3240 header having SATA shelf, so that my FC and SAS remains in one header, while all the SATA is in another header. But those shelves which I am migrating from 3020(having soft disk ownership) and 3040 are having root volumes on them. So when I connect these shelves to 3240 cluster, then each of the headers will see two root volumes, one for its own, and the other for the migrated shelve. So how to handle this situation? I have a long solution, that is creating a small 3 disk aggr on a temporary shelf and add it to 3020 and 3040 one by one, move the root volume to this temporary aggr, and then migrate the original shelf to 3240, but it will be a lengthy process, as each time I have to change ownership of disk, destroy temp aggr , attach them to those filers, do disk zero on these 3 disk of the temp aggr etc, each time.

2. While moving the FC shelves, to the 3240 header with SAS shelf, I have to connect the FC shelf to both filers. Now the issue is my 3240 is having only 2 FC ports per header, and I have two types of FC disks, 15krpm, and 10krpm. So if I have to connect 15krpm disk shelf and 10krpm shelf in one loop, and use the other FC port for multipath in the primary filer, and in the partner, I do not have any shelf, so both the FC ports will be used to connect to the primary loop and the multi path. But my question is, since I am using single loop for my 15krpm shelf and 10krpm shelves, will it have any issues, or what is the best architecture, with the constraints I have.

Thanks in advance.

-Rajdeep

[aborzenkov]

1. You will likely be adding shelves online, in which case nothing bad happens. You will see new foreign aggregate that you simply immediately destroy after that.

2. I doubt that having 10K and 15K on the same loop will have any measurable impact by itself. But 10K are likely 2Gb/s disks and 15K - 4Gb/s disks, which means, you will need to run the whole loop with 2Gb/s. Actually, if FAS3240 runs 8.1, 2Gb/s components may simply be out of support now; you have to carefully check system configuration guide.

[rajdeepsengupta]

@aborzenkov, my 10krpm disk shelves are with ESH4, so I can configure it with 1/2/4 gbps, and it is supported in the 8.1. The DS14MK2 with ESH2 module is not supported, even DS14MK2 AT-FCX is also supported. So supporting is not an issue.

Also can I add DS4243 or DS14MK4 FC disk online, I means is shtdown of the filer not required?

Thanks

[aborzenkov · Accepted solution]

Yes, you can add new shelves online.

[rajdeepsengupta]

Thanks to Aborzenkov, for his reply on the above question.

Just want to summarize

1. From the CP point, if you have SATA, SAS and FC disk, then it is better to keep one header with SAS and FC drives, while the other header with SATA drives. This will make sure that the CP writes are optimized.

Also to add a shelve with root volume to another header with root volume is simple. First move all the volumes on the shelve offline, and then remove the disk ownership. Once done, move the shelve to the running new filer, and change the disk ownership to the new filer(if auto assign is disabled). Then once aggr is online, offline the root volume of the transferred shelve, and destroy it. Then move all the data volumes online.

2. On 3240 if you only have two FC ports, so connect the FC shelves in daisy chain in one port, and use the other port for multi-path. And if you have cluster partner than use the FC ports in the same fashion.