Hi all,
Need some advise about possible way how to go about this. I’ve inherited 2x 8440s that are equipped with 40x 3.89TB SSDs each. (Unfortunately, back then the administrator decided to go for such huge SSDs which are expensive, in any case I’m now stuck with these). The company has now decided to expand this storage but since the SSDs were too expensive the decision taken is to get 1.8TB SAS 10K disks.
Current Setup is :
4x node 8440
2x cages (4 in total including the controller nodes)
40x 3.84TB SSD (10 disks per shelf).
Attachment:
File comment: Existing Layout
existinglayout.png [ 98.92 KiB | Viewed 23472 times ]
The expansion will bring 6x shelves and 20 disks per site. The layout we were thinking of would be :
Attachment:
File comment: After Expansion
afterexpansion.png [ 95.46 KiB | Viewed 23472 times ]
We understood that we can relocated the SSDs so that we achieve balancing across all enclosures but it seems that this is going to be a lengthy process. On the other hand we are thinking on migrating all data from one storage to the other, level one site and once finished migrated data back and do the other storage. We have enough capacity and this storage is still barely used.
What would be the correct procedure to balance the disks across all enclosures (the SSDs) keeping in mind that we don’t care of any data since this will be migrated. Should we just delete the current CPGs , remove disks and redistribute accordingly and re-create the CPGs again ? Is this be possible ? I’ve read somewhere that most probably we will need to re-initialize the whole storage again (contact HPE and re-initialize the storage to its mob).
Also what would be the recommended CPG config + step size if we would say that we are going to allocate 2 or 1 SSD per enclosure for AO while using the other SSDs for throughput hungry hosts and the rest of SAS disks for other applications?
Thank you
Etienne.