• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

ESXi 6 RAID1 support for OS/Datastore?

eduncan911

n00b
Joined
Mar 4, 2012
Messages
15
Does ESXi 6.x has native support for RAID1 and Datastore (rest of the drive) if the proper chipset drivers were to be loaded?

Specifically, the Intel RSTe (for server boards) type "software raid" kind of RAID.

TL;DR

So the story goes that ESXi does not support anything but hardware raid HBAs (like my IBM M1015, and other famous add-on cards and onboard SCSI/SAS controllers), etc etc, blah blah.

After a lot of trolling, I finally found some instructions on building a developer environment to compile my own ESXi drivers. Seems pretty straight forward (I'm versed at this level of Linux).

But before going through that extreme, I wanted to know if ESXi will even support it.

It comes down this I think:

Does the ESXi kernel stack allow booting from a custom driver?

I'm thinking yes because there are custom NIC drivers you can install, right? But I'm afraid the HDD drivers is only what is exposed through AHCI and SAS interfaces native to the kernel modules.

"Hardware Raid" works because the BIOS reports the memory address of the OPTION ROM to the bootloader, which looks like a normal OPTION ROM. It presents itself as a special type of AHCI drive looking like just 1 container. The HBA itself handles the interruption of the RAID containers and metadata overhead of what those containers are: the OS doesn't care.

"Software Raid", such as Intel RSTe in the C612 chipset I am looking to enable, actually has the BIOS supply that metadata required for the RAID set created in a different memory address: the drives are still presented to the bootloader as normal AHCI drives. Therefore, the kernel drivers for the software raid needs to be able to read this metadata. Think of Intel onboard raid acting like an intermediate layer between the kernel I/O operations -> Intel RAID driver -> BIOS AHCI hardware instructions. It's a "software" layer that is interrupted at the OS level, with a driver, that knows to treat the AHCI drives as a RAID1 device (and passing replication instructions to the controller, which then handles the duplication).

Like I said, I am 92.3% sure I can successfully create an ESXi 6 VIB driver package using the CentOS 7 Intel RSTe driver package already available (ESXi is CentOS based, and you need CentOS to build custom drivers for ESXi).

But I'd like to know the likely hood of success before spending that all the time hacking away at it.

Can you load a VIB that acts as an ESXi kernel module?

For this I am thinking "yes" only because of custom NIC driver support you can implement with VIBs. If you are able to side-load a custom VIB into the ISO to get the NIC to be detected, then why not an additional AHCI/RAID driver?


What is my ultimate goal?

I want the ESXi install, and "Datastore" to be on RAID1 using my Intel C612 chipset onboard RAID1 using 2x 240 GB SSDs I have lying around.

I have 2x older 240 GB SSDs sitting here next to me that I'd like to use in my new ultra-lowe powered C612 server build. I want the RAID1 redundancy because, well, they are OCZ SSDs (enough said). I do not want to install an HBA with hardware raid just for the sole purpose of RAID1 P.o.S. drives. I do not want to install ESXi on a USB stick and use just one of the 240 GB drives as a Datastore either.
 
Last edited:
Have you looked at Gea's ZFS all-in-one? Go with ZFS to handle your file system, then share via NFS/iSCSI to ESXi as a datastore. Better support, and a much better file system for data integrity.
 
In short, no, as none of the other parts are there to handle it - ESXi is not linux, it does not have the compatibility software to make it possible to do what you want to do. Certain drivers (not storage) have the ability, for now, to use the vmklinux compatibility layer to function, but that's a deprecated feature and doesn't support storage devices (other than bare-minimum to support the older P400 cards that presented as block devices, not scsi devices).
 
The story isn't a story - it's how it is. You get to spend one of two things: time or money. Your choice. My time is more valuable to me, so I just bought the HBA.
 
If you don't want to buy an enterprise grade Raid 1 SD card adapter then you can get Raid 1 SD card adapters that plug straight into SATA are like $20 and they are (afaik) completely transparent to the OS (even ESXi).
 
Have you looked at Gea's ZFS all-in-one? Go with ZFS to handle your file system, then share via NFS/iSCSI to ESXi as a datastore. Better support, and a much better file system for data integrity.
That sounds decent, except I do plan on some pass-through as well (extending to GPUs and HBAs in the future). That sounds a bit complicated. But interesting.

In short, no, as none of the other parts are there to handle it - ESXi is not linux, it does not have the compatibility software to make it possible to do what you want to do. Certain drivers (not storage) have the ability, for now, to use the vmklinux compatibility layer to function, but that's a deprecated feature and doesn't support storage devices (other than bare-minimum to support the older P400 cards that presented as block devices, not scsi devices).
Thank you. That stops me before wasting my beauty sleep on it in the middle of the night. :)

If you don't want to buy an enterprise grade Raid 1 SD card adapter then you can get Raid 1 SD card adapters that plug straight into SATA are like $20 and they are (afaik) completely transparent to the OS (even ESXi).
Those look pretty cheap, and I more worry about TRIM and drive failure notices.


Thank you all for the replies. In the end, I'll install it on a single SSD and use a backup utility to do a copy to another SSD attached. Looks pretty damn simple too to do that from the Hypervisor itself:

http://33hops.com/blog_xsibackup-rsync-considerations.html
 
It'll freak out on the VMFS datastore - signatures won't match. Boot from a USB stick, it's much saner.
 
So have all that horsepower in my server; but yet, boot from a tiny usb stick just seems... wrong.
 
So have all that horsepower in my server; but yet, boot from a tiny usb stick just seems... wrong.

Why? ESXi doesn't use the boot drive except to boot, and to update the bootbank on a scheduled basis and during shutdown. It runs entirely from a RAMdrive once up and going, and that's pretty much ~it~. Give it a scratch space and a core partition and you're golden.
 
Most modern motherboards have an onboard USB port, or you can get an adapter to use one of the USB header ports. No need to have a USB port hanging off an external port.
 
Back
Top