Suggestions on New FAS Platform

[netappmagic]

Any suggested new platform to replace FAS806 for a 4 nodes cluster? 2 of them are FAS8080

[netappmagic]

Anybody can please pick up this question?

[AlexDawson]

Hi there,

I don't quite understand your question. You have a 4 node cluster, with one pair of FAS8060s and one of FAS8080s? These platforms are both still supported by current versions of ONTAP, and we will provide support contracts through end of 2022 on them.

What challenges are you facing? What does your budget look like? What is your driver for change?

Current platforms are FAS2600(AFFA200), FAS8200(AFFA300) and FAS9000(AFFA700), and are available with spinning disk, SSD or hybrid configurations. We also have an all flash compact high performance unit, the A700s.

[netappmagic]

AlexDawson,

Lease for the paiir of two FAS8060's will be expired, so, we are looking for simiar capacity to these two nodes. So, what wiill be your suggestions?

[AlexDawson]

Ok, I understand.

System sizing aspects include capacity, performance, parallelism, as well as constraints such as budget, space, power and cooling, and enhancements such as support levels and software licensing. This topic is probably too complex and commercially sensitive to cover on our forums.

At a broad stroke, a single HA pair of FAS9000 systems will likely meet your requirements, but without detailed analysis of your environment, we couldn't say that is the correct choice for you.

Your leasing reseller should be able to work with you to identify your sizing requirements for a replacement system.

Please use this form if you would like to contact our sales engineering team for your region - https://www.netapp.com/us/forms/sales-contact.aspx

[netappmagic]

Thanks!

One follow up. Are there any ways to move the aggregate along with all data over to the other pair of HA within the same cluster?

[AlexDawson]

Yes, with Clustered ONTAP you can add a new set of nodes and drives to the cluster and move the data across using "vol move" functionality, just as if you were moving it between the existing four nodes. You will need appropriate cables/SFP+/optical fibre to connect all 6 nodes simultaneously to the cluster switches for migration. Post migration, you would unjoin the departing nodes, and then convert the cluster to switchless clustering. 

Or if you own the disks but not the controllers, we have a "non-disruptive headswap" procedure using aggregate relocation available for our services partners. This is a safe, but complex, activity that we only support for trained technicians.

Hope this helps!

[netappmagic]

For your 2nd option, if I wanted to migrate an aggr including data on it to another node and in the other HA pair, we have to reconnect all shelves associated with aggr to the target node, in this way, I don’t have to run vol move because it will take long time for a large amount of data. Can we achieve that?

I am not seeking a way to do it myself, but just essential steps for what options we can have

[AlexDawson]

As I said earlier - I can't say for sure that a single HA pair FAS9000 is the correct choice - a 4-node FAS8200 cluster might be correct, and then in that case, our ARL headswap procedure would work.

If the disk is also being returned, and replaced like for like (capacity), you could use "storage disk replace" to copy then fail-out the old disks one at a time

[netappmagic]

I need to keep an aggr from one of 2 away nodes. So, my question is if I can move the aggregate over to the other existing HA pair in the same cluster, then without moving data on the aggr.

The following is the doc I found and sounds like it can do the work. However, I DO recognize that this doc is for 7-mode, not for CDOT. Are there any precedures we can do to achieve the same for CDOT cluster?

https://library.netapp.com/ecmdocs/ECMP1196986/html/GUID-85C8CF29-1FF6-4DC1-9835-EB89E7EA1BF6.html

[AlexDawson]

Moving aggregates between systems when not performing headswap upgrades is not supported on Clustered ONTAP as it was with 7-Mode.

Also, your comment mentions "away nodes" - is this a metrocluster system? If so, you really need to work with a local partner or NetApp SE team for planning this migration.

[netappmagic]

OK.

Can we do "headswap upgrade" if HA pairs will be replaced are two FAS nodes, and two new nodes are AFA's ? 

If the answer is "NO", then we have to use the method as you described previously that connect all 6 nodes simultaneously to the cluster switches for migration. Correct?

It is not metrocluster. 

[AlexDawson]

AFA (AFFA) systems will only accept solid state drives, so unless all of the previous drives are SSD, it would not be possible to perform non-disruptive aggregate relocation (ARL) headswap onto one.

ARL Headswaps must be 1 to 1, so all drives off one system must go to another one, and only drives from one source system can go to a destination system.

If this does not suit, then yes, you would need to have all 6 nodes connected and then perform vol moves.

[netappmagic]

One more follow-up, if you don't mind.

Can headswap happen between two existing HA pairs in the same cluster? In another words, move all existing aggrs from one HA pair to the other that has had its own existing aggrs, and then retire the HA pair?  Assuming they are both FAS nodes, and accepting HA can accept enough shelve.

[AlexDawson]

Sorry for the delay in response. Our procedure for aggregate relocation does not cover a many to one scenario, only a one to one.

From a technical point of view, it might work - but from a practical point of view, if anything went wrong, it could be a bad scenario for you. It is not something documented in our procedure for aggregate relocation, which assumes the existing target does not have any disks - specifically saying "This procedure assumes that the replacement controller hardware is new and has not been used."

For the safety of your data you should assume it will not work and pursue consolidation options through vol move, potentially onto swing disks, or do 2x8060HA to 2x8200HA. 

[aborzenkov]

There is no documented way to move shelf with data between nodes in cDOT. Option 2 is about replacing nodes themselves. Nodes fully retain their identity in cluster, their shelves still continue to belong them.