unRAID in ESXi - using a virtual drive for cache/array drives

I wasn't really sure what section to post this in - this seems the closest fit.  I wonder if it's worth us getting a new 'Virtualization' (it was really difficult for me to spell that with a Z ) section on these boards, especially considering the rising popularity of ESXi-based unRAID servers.

In any case - there has been a long-standing issue with unRAID whereby when you try and add a virtual disk (vmdk) as a source for a cache drive (or any drive for that matter), you can't assign it to the array when using any of the SCSI adapters.  If you create a vmdk using an IDE adapter, it works, but is horribly slow.  Why would you want to do this?  Say you want a cache drive on a superfast array that is also internal to your host, so you can leverage the benefits of the 10Gbps VMXNET3 adapter and exceed the constraints of physical 1GB NIC's. 

To get around this issue, try using the VMware Paravirtual controller in order to leverage the pvscsi driver that was introduced in unRAID 5.0rc5, in combination with the following settings. 

Instructions follow:

• Shut down your unRAID VM
  
• Right click the VM in vSphere client and choose "Edit Settings"
  
• Highlight the existing SCSI controller (if you are using a vmdk boot method, it is probably 'LSI Logic Parallel') and click 'Change Type'.  Choose 'VMware Paravirtual' from the list and click OK
  
• If you aren't using the vmdk boot method, then simply add a new virtual drive to your VM and choose 'VMware Paravirtual' when prompted for the controller type.
  
• Still in the 'Edit Settings' menu, click on the Options tab, then highlight 'General' under the Advanced section.
  
• Click 'Configuration Parameters' at the bottom right and then click 'Add Row'
  
• Add the 'disk.EnableUUID' parameter as 'true', so it looks like the following screenshot.
  

That's it!  Click OK a couple of times and start your unRAID VM.  You'll now be able to add SCSI vmdk's as drives the array manager.  Just be careful you don't inadvertently add the vmdk boot drive to your array.  If you're not sure which one that is, you can start the array first, note the disk reference for the one that is smaller in size (eg sdb 1.0GB) and not add that one.  That's the only real downside to this method, in that you can't easily identify the size of the vmdk you are about to mount until it is mounted.

Thanks BetaQuasi

I've been working on setting up unRAID 5.0rc12a as a VM.  With your vmdk image file it was pretty easy.  I got hung up on trying to getting the drives visible in unRAID.  I originally configured the ESXi SCSI controller as "LSI Logic Parallel", with no luck.  Then I read on another that someone switch theirs "LSI Logic SAS".  unRAID was now to seeing the drives in the array!!   

When I used the PVSCSI or "LSI Logic SAS", unRAID's main menu shows the all drive devices spun down.  Flashing green.  Which is false, for I can read and write to the array.  With them flashing, I cannot get "TEMP" readings.

Do you loose this feature when you use a virtual SCSI controller? 

VMware PVSCSI driver - version 1.0.2.0-k
vmw_pvscsi: using 64bit dma
vmw_pvscsi 0000:0b:00.0: vmw_pvscsi: host->max_id: 16
vmw_pvscsi 0000:0b:00.0: irq 74 for MSI/MSI-X
vmw_pvscsi: using MSI-X
scsi0 : VMware PVSCSI storage adapter rev 2, req/cmp/msg rings: 8/8/1 pages, cmd_per_lun=64
vmw_pvscsi 0000:0b:00.0: VMware PVSCSI rev 2 host #0
scsi 0:0:0:0: Direct-Access     VMware   Virtual disk     1.0  PQ: 0 ANSI: 2
scsi 0:0:1:0: Direct-Access     ATA      ST3000DM001-1CH1 CC24 PQ: 0 ANSI: 5
scsi 0:0:2:0: Direct-Access     ATA      ST3000DM001-1CH1 CC24 PQ: 0 ANSI: 5
scsi 0:0:3:0: Direct-Access     ATA      ST3000DM001-1CH1 CC24 PQ: 0 ANSI: 5
scsi 0:0:4:0: Direct-Access     ATA      ST3000DM001-1CH1 CC24 PQ: 0 ANSI: 5
scsi 0:0:5:0: Direct-Access     ATA      WDC WD5000KS-00M 07.0 PQ: 0 ANSI: 5

How are you connecting your drives?  Virtual controllers are not relevant to physically passed through controllers, and you should still get temps if you are passing through a controller.

RDM'ed drives behave differently though - you won't get temps and often spinup/spindown doesn't work correctly either.

I have the drives connected directly to the on board SATA ports.  I then created a RDM vmdk file of each disk with the -z option instead of the -a.

I figured I had done something wrong, since the spinup/spindown was not working as I had expected, if that's normal, then I am not going to worry about it.  What would recommend to use to monitor/track the vitals of a ESXi system?

Thanks!

I have the drives connected directly to the on board SATA ports.  I then created a RDM vmdk file of each disk with the -z option instead of the -a.

 

I figured I had done something wrong, since the spinup/spindown was not working as I had expected, if that's normal, then I am not going to worry about it.  What would recommend to use to monitor/track the vitals of a ESXi system?

 

Thanks!

I have the same issues.  When I used -z I was able to access the SMART information.

I tried the hdparm -y and it does not work.

I'm wondering if I should reboot directly into unRAID, then use the hdparm -S242 command so that the internal drive timer will work.

I.E. Turn off unraid's spin down and let the drive timer do the work.

I'm not sure if ESX will allow that. I've noticed it probes the drives every now and then.

I could use some help on this as well. I'm recently getting start with ESXi 5.1 and unraid. I used this command:

vmkfstools -a lsilogic -z /vmfs/devices/disks/t10.ATA_____ST3160812AS_________________________________________5LS3P8SB   mydisk1.vmdk

(physical name different in my machine)

From the post over here http://lime-technology.com/forum/index.php?topic=7914.0

I have tried both the SCSI Logical SAS and the paravirtualized controller methods. So far I'm not getting any physical info from the disk and spin control is missing.

In this case my test mobo is an Asus P5Q Deluxe with the disks on the ICH10 6 port SATAII controller. If I ever get this working and I get confident my production mobo will be different. I'd like some sense of what will work. Not spinning disks down is not an option.

Thanks

If not spinning down disks is not an option, then RDM'ing drives is not an option - I answered this in the other thread.

You need to get a SATA controller (M1015, SASLP-MV8) and pass it through to the VM if you want disk spindown.

I think I may do this to create an additional HD (a second VMDK) that I can mount outside of the array.  This would be the perfect place to install plugins, webroot, etc.  And, when it comes time to update unRAID to the latest RC (or final), all I would need to do it replace the primary unRAID VMDK and my plugins won't be messed with.

Sound like it will work?

John

You need to get a SATA controller (M1015, SASLP-MV8) and pass it through to the VM if you want disk spindown.

I installed a SASLP-MV8 and have it configure setup as a pass-through to unRAID.  I then added a cache drive and bingo!  I have the get spin up/spin down and temp readings.

I will look at moving all the unRAID drives from the on-board SATA ports to the SASLP-MV8.  I don't think there will be a performance hit.

Thanks BetaQuasi!

I think I may do this to create an additional HD (a second VMDK) that I can mount outside of the array.  This would be the perfect place to install plugins, webroot, etc.  And, when it comes time to update unRAID to the latest RC (or final), all I would need to do it replace the primary unRAID VMDK and my plugins won't be messed with.

 

 

Sound like it will work?

 

John

Sounds like it would work for sure - and a great idea to boot.

I know I'm bumping an old thread, but a request to move this to the "unRAID as a Guest" forum.