Best way to move customer from smaller pools to larger?

[adamdb_uk]

Hi

 Just wanting to confirm my logic for the following plan for an MSA2050

Customer has 2x RAID5 3+1 disk groups (one pool a, the other pool b). 

Customer wants to fully populate array with same 600gb sas drives but end up with

2x RAID 5 9+1 disk groups in each pool plus 2 spares per pool.

Currently bays 1-8 are in use - no spares.

My plan

1. Create new DG in pool in slots 10-20 RAID5 9+1 in pool b

2. Remove existing 3+1 RAID5 DG in pool B forcing data migration to new DG in pool B

3. Using ESXi storage vmotion move data from pool A into new larger pool B.

4. remove old DG in pool A and recreate as 9+1 RAID 5 in slots 1-10 (using the 8 original drives and 2 new ones).

5. migrate data back into pool A from pool b using storage vmotion.

5. add two spares per pool in slots 21-24.

This means hopefully no downtime for the customer but will mean some performance impact.

It's either that or delete both DG's and start over  re-creating the DGs with restoration from backup.

Hopefully I've not missed anything.

Any feedback appreciated. I've not touched an MSA for a while.

thanks.

[JonPaul · Accepted solution]

@adamdb_uk 
Your process is sound there are a couple things I would make you aware of:
1. Your proposed final configuration, RAID 5 9+1, does not match the best practices and could impact your Sequential WRITE performance.  See MSA Best Practices and MSA Virtual Storage Technical Guide
2. Disk-Group DRAIN will take much longer than you think it should.
DRAIN uses the tiering engine to migrate data which is designed to not impact standard I/O, i.e. low priority.  Based on the current size it should not be too terrible but still would plan in the DAYS range to get the disk-group DRAINed.
Lastly, the best practice recommendation would be to use RAID 6 with spinning media.

[adamdb_uk]

Hi,

      thanks for the update. Yes I'm aware the drain can take some time to complete. To be honest I've never tried it before so wanted to make sure my understanding of how it works was correct. With regard to the raid configuration I'm aware R6 is normally the best practice for spinning media  however this particular customer has priorotised capacity over performance/resiliance and has already gone this route elsewhere at other sites. I will also look to uplift controller firmware to the latest version before attempting any works (I believe they are running 2019 code which is now deprecated)...

thanks for flagging up the issues and at least confirming the steps are valid. Appreciated.