VMware Cloud Community
spaceinvader
Contributor
Contributor

host didn't switch off ("Requesting system poweroff")

I've an ESXi 4.0 Update 1 host on a Supermicro H8DIi+-F mainboard.

If I want to power off the host (via vSphere Client or via host console) it ends up in "Requesting system poweroff" (on Alt-F1 console). But the host don't poweroff.

(With an alternative OS (Windows) on the same machine the shutdown is working correctly.)

The BIOS' ACPI settings seems to be ok, too. (Too much to report all of them here. I.e. "ACPI aware OS" = "yes")

(I remember in the beginning of my tests the host was powered off correctly - but I'm not sure. (I'd meanwhile: a strong crash after adding a PCIe card and driver - forced to load BIOS defaults, and update ESXi 4.0.0 to Update 1.)

Does somebody have a hint where I've to look for?

Thanks!

Reply
0 Kudos
16 Replies
a627tqt
Contributor
Contributor

So you got it successfully installed with the BIOS defaults? Mine keeps hanging up at the yellow screen during install "Starting vmkernel initialization". I've changed BIOS setting at least 20 times enabling /disabling USB, ACPI, etc...

Any suggestions?

Reply
0 Kudos
spaceinvader
Contributor
Contributor

I've sucessfully installed ESXi 4.0 (later updated to u1) on a H8DiI-F and H8Di3-F.

With default BIOS settings, and with many different BIOS settings, too.

But I'm using (for VMware OS partitions) HDDs at Adaptec RAID controllers, or partitions on USB sticks.

The VMFS datastore partitions I've placed on HDDs at Adaptec RAID controllers, or at onboard SAS controllers (of the H8Di3+-F).

The onboard SATA controller seems to be recognized as an "SB700" (and not precisely that one which is integrated into the chipset / SP5100 chip, perhaps a similar successor of the SB700). Additionally only four of the six SATA ports are available with this driver.

(I don't trust in 100% reliability of the onboard SATA controller usage, therefore.)

(By the way: we don't have to talk about PATA. That's outdated. Disable it.)

But I've used an additional HDD at the SATA controller for backing-up/transferring VMs (formatted as VMFS, too). All of this was done by vSphere Client techniques.

(So it could be possible that initial installation to HDD at SATA controller is working, anyway - but I didn't tested that ever.)

(Be sure "OnChip SATA Type" is "Native IDE" (BIOS default) <- Advanced Chipset Settings, SouthBridge Configuration.)

Perhaps you're choosing one of the not supported two SATA ports? SATA0 should be recognizable (is "third master", the 5th of all 8 IDE).

spaceinvader

Reply
0 Kudos
DSTAVERT
Immortal
Immortal

On questions about unsupported hardware you are more likely to find answers at http://vm-help.com.

-- David -- VMware Communities Moderator
Reply
0 Kudos
a627tqt
Contributor
Contributor

yeah, I have a different issue on the H8DI3. I don't even get to the point of selecting drives. It freezes and I lose the keyboard as soon as the "Starting vmkernel initialization" yellow screen. I've tried noacpi, etc... but still no luck. I did find some info on processor stepping being different that may cause this. I'm going to investigate that a little but I'm getting close to boxing it up and sending it back.

Reply
0 Kudos
DSTAVERT
Immortal
Immortal

Trying to make something work that isn't on the Hardware Compatibility List is difficult. If this is production then I would not stray from the list. If this is for testing and exploration then consider installing ESXi inside VMware workstation.

-- David -- VMware Communities Moderator
Reply
0 Kudos
a627tqt
Contributor
Contributor

it wouldn't be so difficult if there was a way to see verbose logs of what's going on. I can't even seem to get to the altf12 or altf1 screens to see why there's an issue. I understand the HCL, but things get added to it all the time, and it's not like this is crappy hardware. if it's truly incompatible I'll return it, but I'd love to be able to see what's going on

Reply
0 Kudos
DSTAVERT
Immortal
Immortal

It is up to the manufacturer to certify the MB/System/Component. Supermicro is well represented on the HCL. Why choose something not on the list?????

-- David -- VMware Communities Moderator
Reply
0 Kudos
a627tqt
Contributor
Contributor

we had originally purchased this server to be deployed as a SQL server, but another came available and this one had more than enough horsepower for esxi. 2x 6 core opterons, 16GB ram, etc... Just seemed like a good fit.

Reply
0 Kudos
spaceinvader
Contributor
Contributor

By the way: I used only one quad core Opteron and only one memory slot group populated, of course.

Reply
0 Kudos
spaceinvader
Contributor
Contributor

It is up to the manufacturer to certify the MB/System/Component. Supermicro is well represented on the HCL. Why choose something not on the list?????

Because someone want to choose boards that fits to his demands! (Not all customers are interested in using prefabricated high price and low equiped solutions of others. - But, of course, they have some risks for trouble...)

Chosing 100% VMware certificated solutions is for unskilled workers, "setup droids", or similar. Smiley Wink

And this board is running with ESXi 4 in general. My systems are working.

The AMD chipset SR5690/SP5100 is developed especially for virtualization solutions. But it's new. So you have to calculate running into problems.

spaceinvader

Reply
0 Kudos
spaceinvader
Contributor
Contributor

we had originally purchased this server to be deployed as a SQL server, but another came available and this one had more than enough horsepower for esxi. 2x 6 core opterons, 16GB ram, etc... Just seemed like a good fit.

With hex core Opterons I would assume problems. (But it's only based on a feeling.)

(But I would try it with one Opteron, anyway. - The supplier will be happy if he gets back systems where do-it-yourself mechanics fiddeling with CPUs... Smiley Wink )

spaceinvader

Reply
0 Kudos
a627tqt
Contributor
Contributor

turns out it's bad memory. with 16 dimms it's going to take a while to track down, but that's easy. Thanks for the assist. I actually discovered it by booting memtestx86+ and it hung up almost immediately! pulled a bunch out, rebooted and esxi boots!

Reply
0 Kudos
DSTAVERT
Immortal
Immortal

It is not uncommon, and perhaps could be the norm, that certification is done with a single processor.

When you take the component approach to providing a server, you ultimately become the manufacturer. When IBM, Dell, HP, Fujitsu, SUN et al. manufacture a server they also get components from suppliers. The difference is that the suppliers and the manufactures get together to make sure the components work well together and keep working well. You as a manufacturer don't have the resources to make sure things work well together and to keep things working well when issues are uncovered.

I use HP servers and they regularly release Firmware updates specific to my particular servers and my version of ESXi. The Firmware can be motherboard, baseboard, or drive controller related but can even extend to the hard drives. All the updated components are tested to work together.

ESX(i) is very demanding of hardware. The supported list is short by comparison to general purpose OS's. When you move from a single OS per server to many virtual servers per host it becomes more important that everything be right.

-- David -- VMware Communities Moderator
Reply
0 Kudos
spaceinvader
Contributor
Contributor

turns out it's bad memory. with 16 dimms it's going to take a while to track down, but that's easy. Thanks for the assist. I actually discovered it by booting memtestx86+ and it hung up almost immediately! pulled a bunch out, rebooted and esxi boots!

I'm curious:

Are you sure you've bad memory modules?

Or may it be a problem of fully populated slots? (With 16 modules there are special issues, like power consumption and line amplifiers / buffers etc..)

(I would expect problems, even if it should run corresponding to technical facts sheets. Especially with cheap or mixed modules.)

spaceinvader

Reply
0 Kudos
spaceinvader
Contributor
Contributor

Hi DSTAVERT!

Of course you're right!

And of course as a "micro micro manufacturer" I've only small resources - and almost no support by component and software manufacturers (which are only interested in the big money).

But (by the way) if I've no resources, I should draw myself back to a snail shell as an unemployed loser.

If I only be able to re-sell HP/DELL systems (spoken in general, not an attack to you!), I would be an unskilled salesman, and this is only a boring job, and can be done thousands times better by a person with low education / low engeneer skills.

That's too few for me. Smiley Wink

But there were several situations where I'd choosed HP servers, too. (Unfortunately I don't like their poor RAID controllers and their exorbitant high price HDDs (especially because they are the same re-labeled models from the well-known HDD manufacturers - higher prices are ok, but not four times higher).)

edit: Without pioneering spirit we would still live in caves.

spaceinvader

Reply
0 Kudos
DSTAVERT
Immortal
Immortal

It only takes one small failure with an unsupportable piece of hardware and the costs far outweigh any "savings" you might believe you make. Just think about the time you have spent just making your server work and now something doesn't. If you manage the system remotely think about this when you need to drive to the server when you need to restart it rather than knowing that it will restart itself. Those are the times when you could be doing something that made you money. I have made my share of servers over the years and I can tell you that as soon as I started getting real servers my self support costs dropped dramatically.

I am not a big shop with bags of money but that is precisely why I will not take chances with the platform that makes me money. If you have nothing to loose then you have no reason go the route I take. All the server manufacturers have deals on returned merchandise or refurbished models. There are severs with fewer frills. You can lease them. Get customers to finance them. Lots of creative was.

BTW the pioneers are the ones who die early after a hard life. Smiley Wink

-- David -- VMware Communities Moderator
Reply
0 Kudos