So, what's the word on passing this through on esxi 5.1 with a I/O MMU supported host? I have yet to try it, but will be needing the extra ports when I finalize the move.

 

I've heard it's possible, but what kind of hackery is required?

And, what's the perfomance?

Does it support spin down?

How about SMART data?

 

Look in the Atlas thread.  Johnm has links to the actual Hack that makes it possible to pass through.  You have to edit the passthru.map file and the VMX file for the VM to complete the hack.  I don't use the same method for editing the VMX file however.  You can add the same entry through the Web Client GUI from Windows.  If you want I can find my post where I document the online edit.  Since you are passing through the controller to the VM it is like it would be with bare metal usage.

Same problem with AOC-SAS2LP-MV8?

I don't know myself.  I've seen a post that suggests it is necessary but I do not have a card so I can't confirm.  I did do the hack on a card that didn't need it.  It didn't seem to break anything but there is no guarantee that it won't break it when used when it shouldn't be either.

In my experience, you've always better off with LSI.  Pretty much every server on vmware's HCL uses LSI for storage.  Hacks are all fine and good but do you really want to hack your storage controllers into working?

I went through the steps of 'hacking' the MV8 into my unraid server, and while it worked ... i'm still slightly hesitant with this.

 

I mean, after all, if it was 100% stable, I'm sure Vmware would have supported it. The lack of support makes me inclined to think there is some unknown issue, and unknown issues cause data corruption.

 

I did some testing, and the MV8 in passthrough is only about 10% faster than onboard SATA for a AMD 990FX mobo. Not a huge difference. I should test it w/ MV8 in non-passthrough mode. If it's 10 or even 20% difference? Probably not big enough for me to take a risk with the data.

 

I mean, after all, the reason we're building these large machines is to protect the data, right? Sacrificing known protection for a 10 or even 20% increase in speed? ... sounds foolish to me.

 

Either way, I'll do a 3 pass pre-clear on these last two drives ( 3 year old Samsung HD154ui) ... Do a dd benchmark, then skip passthrough and pre-clear & DD benchmark them again.

 

 

I went through the steps of 'hacking' the MV8 into my unraid server, and while it worked ... i'm still slightly hesitant with this.

 

I mean, after all, if it was 100% stable, I'm sure Vmware would have supported it. The lack of support makes me inclined to think there is some unknown issue, and unknown issues cause data corruption.

Some of this is indicated with my Windows 7 VM incompatibility: Even with the Hack I get BSODs setting up or using Windows 7 with a MV8.  It does however work fine in WHSv1 or XP so could just be a Windows driver issue.  The VMX edit turns on IOAPIC for the VM.  Which changes how interrupts are handled according to following document.  This is from 4.0 but should be same for 5.0 & 5.1: http://www.vmware.com/files/pdf/techpaper/vsp_4_vmdirectpath_host.pdf. 

My experiences so far?

 

My stock system (FX8350 + 32gb Ram + GIGABYTE GA-990FXA-UD3 AM3+ ... utilizing the hacks provided on my 5.1 799733 server led to nothing but difficulties. With the hack in place (with SASLP physically installed or not) I got nothing but purple screens and hardlocks.

 

Was this because of my stock firmware on the SASLP? Or perhaps I did something wrong on the hack? I dunno.

 

My experiences may not be normal ... but it's what I saw.  Before I did the hack, I had a zero issues with stability on uptime with 100% cpu usage on VMS, after? Many hardlocks, some purple scrrens. After I removed it? Zero issues. (prior to undoing the customizations on the two files, I'd had three hard locks doing 100% cpu usage)

 

I'd prefer to err on the side of caution here, so, I just ordered another M1015. Especially considering it only cost me a 115.00 (yea, twice as much as 6 months ago, but still well worth the 115).

 

And I've got another 3 1.5TB drives I can add to array, not including the donor drive for data recovery.

 

Ultimately: I can only say this. Test, Test Test for anyone considering the SASLP hack. With the hack, 100% cpu equaled hard lock or purple screen ... (wish I had taken pics of the purple screen) ...

 

Don't do it without a lot (and I mean a lot) of testing.

 

 

I went through the steps of 'hacking' the MV8 into my unraid server, and while it worked ... i'm still slightly hesitant with this.

 

I mean, after all, if it was 100% stable, I'm sure Vmware would have supported it. The lack of support makes me inclined to think there is some unknown issue, and unknown issues cause data corruption.

Some of this is indicated with my Windows 7 VM incompatibility: Even with the Hack I get BSODs setting up or using Windows 7 with a MV8.  It does however work fine in WHSv1 or XP so could just be a Windows driver issue.  The VMX edit turns on IOAPIC for the VM.  Which changes how interrupts are handled according to following document.  This is from 4.0 but should be same for 5.0 & 5.1: http://www.vmware.com/files/pdf/techpaper/vsp_4_vmdirectpath_host.pdf.

Heh, right now I've got 2 hours of 100% cpu usage doing an encode, and it hasn't hardlocked.

 

This makes me a happy camper.

My experiences so far?

 

My stock system (FX8350 + 32gb Ram + GIGABYTE GA-990FXA-UD3 AM3+ ... utilizing the hacks provided on my 5.1 799733 server led to nothing but difficulties. With the hack in place (with SASLP physically installed or not) I got nothing but purple screens and hardlocks.

 

Was this because of my stock firmware on the SASLP? Or perhaps I did something wrong on the hack? I dunno.

 

My experiences may not be normal ... but it's what I saw.  Before I did the hack, I had a zero issues with stability on uptime with 100% cpu usage on VMS, after? Many hardlocks, some purple scrrens. After I removed it? Zero issues. (prior to undoing the customizations on the two files, I'd had three hard locks doing 100% cpu usage)

 

I'd prefer to err on the side of caution here, so, I just ordered another M1015. Especially considering it only cost me a 115.00 (yea, twice as much as 6 months ago, but still well worth the 115).

 

And I've got another 3 1.5TB drives I can add to array, not including the donor drive for data recovery.

 

Ultimately: I can only say this. Test, Test Test for anyone considering the SASLP hack. With the hack, 100% cpu equaled hard lock or purple screen ... (wish I had taken pics of the purple screen) ...

 

Don't do it without a lot (and I mean a lot) of testing.

M1015 is my card of choice too.  I no longer use my SASLP-MV8s since most of my VMs are Windows 7 now.  I only have one VM that could use the MV8 and I just used my spare M1015 for that.  I plan on getting a couple more spare M1015s sometime but haven't gotten around to it yet.  When I had my MV8s installed I had both the 15n and 21? firmwares and didn't really notice a difference.  I've never updated the firmware on my MV8s just ran them stock.