Using RAID in HBA mode or remove RAID?

[aaralves]

Hi,
I am installing ProxMox + CEPH on top of Dell PowerEdge R640 with PERC H740P Mini embedded.
Does anyone help me deciding if I must use the PERC in HBA mode or if it is preferable just disable the PERC?
Tk.

 

[david_tao]

I recommend use PERC in eHBA mode is because beside with CEPH, you will need another boot disk separated from OSD and that is good to use RAID 1 to protect it.
Following Dell document can be referenced for how to create eHBA on H740p.
https://www.dell.com/support/kbdoc/...-create-virtual-disk-in-ehba-mode-for-perc-10

 

[aaralves]

I recommend use PERC in eHBA mode is because beside with CEPH, you will need another boot disk separated from OSD and that is good to use RAID 1 to protect it.
Following Dell document can be referenced for how to create eHBA on H740p.
https://www.dell.com/support/kbdoc/...-create-virtual-disk-in-ehba-mode-for-perc-10

Thank you for your help.

 

[aaralves]

Thank you for your help.

Hi David,
One more question: Do you think that, for the ProxMox boot disk, RAID 1 virtual disk configured at H740p is better than a mirror ZFS?
Thank you

 

[david_tao]

Hi aaralves:
I think both are ok if I know which method I choose. If I use PERC's eHBA to build VDs (RAID1) for boot disk, then if an physical disk in the VD failed, I only needs to pull it out and push a good disk back to same slot then all is done, H740p will help to rebuild all data on it(but one thing is it will write every blocks to the new disk, that is not really good thing for SSD or NVMe). if I use zfs to make RAID1 for boot disk, after I replaced this failed disk by good one, then I will need use following command to trigger ZFS do the fix task(but as ZFS does, Only those blocks that actually contain data will be copied to new disk, that is good for SSD or NVMe). just FYR.
# zpool replace -f <pool> <old-device> <new-device>
And for how to replace failed disk in ZFS, you can reference the ZFS administration portion
https://pve.proxmox.com/pve-docs/chapter-sysadmin.html#_zfs_administration

 

[aaralves]

Hi aaralves:
I think both are ok if I know which method I choose. If I use PERC's eHBA to build VDs (RAID1) for boot disk, then if an physical disk in the VD failed, I only needs to pull it out and push a good disk back to same slot then all is done, H740p will help to rebuild all data on it(but one thing is it will write every blocks to the new disk, that is not really good thing for SSD or NVMe). if I use zfs to make RAID1 for boot disk, after I replaced this failed disk by good one, then I will need use following command to trigger ZFS do the fix task(but as ZFS does, Only those blocks that actually contain data will be copied to new disk, that is good for SSD or NVMe). just FYR.
# zpool replace -f <pool> <old-device> <new-device>
And for how to replace failed disk in ZFS, you can reference the ZFS administration portion
https://pve.proxmox.com/pve-docs/chapter-sysadmin.html#_zfs_administration

Hi David
I think that I will keep ZFS RAID 1 for the debian OS and the CEPH for the ProxMox distributed storage.
Thank you for your comments and suggestions.