Knowledgable
Enthusiast
Enthusiast

vSAN Stretched Cluster - Local and Remote Protection

Jump to solution

Gents,

I have two identical servers with 2 SATA SSDs and 5 x 1.2 TB disks for capacity (layout attached).

- Both side A and B of each server has RAID5 configured at the hardware level

- Each site has one (1) server and a third site hosts the witness.

What I would hopefully like to achieve, is having vSAN objects replicated between side A and B on each server (so i can lose a disk group and still be functional), as well as remote protection at the secondary site.

Hope that was clear. Also attached the storage policy i came up with for review.

0 Kudos
1 Solution

Accepted Solutions
TheBobkin
VMware Employee
VMware Employee

Hello Knowledgable,

"I should not let the RAID controller handle the drive protection from the hardware level and allow vSAN to manage it to be supported ?"

No, they shouldn't be RAID5 before being presented to ESXi/vSAN.

Check the vSAN-HCL entry for the controller - the most accurate way to determine the model is to get the VID:DID SVID:SDID using #vmkchdev -l |grep vmhbaX  , then you can find it via the HCL page or via https://hcl.captain-vsan.com/ .

The entry for this controller will state what controller modes are supported (e.g. RAID0 and/or Passthrough(AKA HBA-Mode)).

Check the current controller mode and how the disks are presented via A hardware-vendor utility (e.g. PERCCLI for Dell, HPSSACLI for HPE) or via out-of-band management console (e.g. iDRAC, iLO, KVM) and ensure they are using the correct mode.

"Also, when i say "remote protection", basically a clone of data on drive bank A on to drive bank B."

This is not possible as each node is a Fault Domain here and it would breach compliance with storage policy to place two RAID1 sets of data components on the same node (this is only possible with multiple nodes per site using PFTT/SFTT as per the link in my last comment).

Bob

View solution in original post

0 Kudos
4 Replies
TheBobkin
VMware Employee
VMware Employee

Hello Knowledgable,

"- Both side A and B of each server has RAID5 configured at the hardware level"

Do you mean the disks are RAID5'ed in a group before being presented to vSAN? (unsupported)

Or that the controller 'personality' is RAID5? (also unsupported, only RAID0 and Passthrough(HBA) are supported and which/both depending on the controller and configuration).

"What I would hopefully like to achieve, is having vSAN objects replicated between side A and B on each server (so i can lose a disk group and still be functional), as well as remote protection at the secondary site."

Normal FTT=1 with RAID1 (a PFTT=1, SFTT=0 Storage Policy in a stretched cluster) will ensure there is a data component replica at each data site. However remote protection (e.g. local FTT=1) is not possible with a 1+1+1 stretched configuration as you have here (this would require a 3+3+1 configuration at minimum).

More information on typical configurations, their FTT, FTM, % used and requirements:

http://www.yellow-bricks.com/2017/05/30/sizing-vsan-stretched-cluster/

Bob

0 Kudos
Knowledgable
Enthusiast
Enthusiast

Thanks Bob,

Just to ensure i follow you, is it that I should not let the RAID controller handle the drive protection from the hardware level and allow vSAN to manage it to be supported ?

Also, when i say "remote protection", basically a clone of data on drive bank A on to drive bank B.

0 Kudos
TheBobkin
VMware Employee
VMware Employee

Hello Knowledgable,

"I should not let the RAID controller handle the drive protection from the hardware level and allow vSAN to manage it to be supported ?"

No, they shouldn't be RAID5 before being presented to ESXi/vSAN.

Check the vSAN-HCL entry for the controller - the most accurate way to determine the model is to get the VID:DID SVID:SDID using #vmkchdev -l |grep vmhbaX  , then you can find it via the HCL page or via https://hcl.captain-vsan.com/ .

The entry for this controller will state what controller modes are supported (e.g. RAID0 and/or Passthrough(AKA HBA-Mode)).

Check the current controller mode and how the disks are presented via A hardware-vendor utility (e.g. PERCCLI for Dell, HPSSACLI for HPE) or via out-of-band management console (e.g. iDRAC, iLO, KVM) and ensure they are using the correct mode.

"Also, when i say "remote protection", basically a clone of data on drive bank A on to drive bank B."

This is not possible as each node is a Fault Domain here and it would breach compliance with storage policy to place two RAID1 sets of data components on the same node (this is only possible with multiple nodes per site using PFTT/SFTT as per the link in my last comment).

Bob

View solution in original post

0 Kudos
Knowledgable
Enthusiast
Enthusiast

Got you, thanks for the explanation.

0 Kudos