High Availability with NFS storage or Ceph with HA ?

[mada]

Hi,

i'm looking to do Storage VM and asking to built 3 or 4 proxmox nodes ( will be up to 17 nodes in future) and looking to got as last 40tb space so my question is what is better choice for me?

i have 10G switch and a lot of 4tb SAS & SATA enterprise drive and some of SSD drives. and around 5 of Raid LSI 9266-4i 1GB Cache.

i read a lot that i can't use normal SATA enterprise drives + LSI raids with Ceph so the HA with NFS will be better options for me?

i'm looking for high i/o speed as well with at last 700MB/s - 1.0Gbp/s

Thanks

 

[fireon]

Never using NFS. It is not a good choice. I thing Ceph is on of the best things for clustering. We have here a littel Cephcluster with 3 nodes. Two 10GB Switches for Ceph and is working perfectly. The easiest way is to use always the exact same model of servers same HDD's and same count of HDDs in every server. We do that, but it is not absolutely necessary.

Importend is that Ceph (same as ZFS) see the controller. No firmware no raid, no controller cache. https://www.proxmox.com/en/training/video-tutorials/item/install-ceph-server-on-proxmox-ve

 

[mada]

Never using NFS. It is not a good choice. I thing Ceph is on of the best things for clustering. We have here a littel Cephcluster with 3 nodes. Two 10GB Switches for Ceph and is working perfectly. The easiest way is to use always the exact same model of servers same HDD's and same count of HDDs in every server. We do that, but it is not absolutely necessary.

Importend is that Ceph (same as ZFS) see the controller. No firmware no raid, no controller cache. https://www.proxmox.com/en/training/video-tutorials/item/install-ceph-server-on-proxmox-ve

Thanks very much so i can use the SAS or the SATA as OSD and SSD as journal?
Something like per node
6 x 4TB SATA & SAS enterprise drives
2x SSD one for promox data and one for Journal

This will be good setup? Or do i need 2x SSD Journal?