VMware ESX 2.5.1 on DL585

[Malige]

Hello,
we have a big problem with this configuration. We use a DL585, 4x2,4GHz,24GB RAM, 2 add NIC NC7771 (Slot 2 and 4) and 2 FC Emulex 9002L Slot 6 and 8 (SAN). No Problems at installtion. But when we install a VM (on SAN or local HDD) the ESX freezed completed, only a cold reset via RILO is possible. We open a call at VMware but all answer didnt help. Now we have to use memtest. Any ideas? Thanks

[Robbert Hoendervanger]

Hi As far as i know VMWARE ESX is not (yet) supported on the DLxx5 Opteron servers.

Q4 this year it will.

[David Claypool]

DL585 *is* on current certification list: http://h18004.www1.hp.com/products/servers/software/vmware/hpvmwarecert.html

[Please, people, quit guessing and posting erroneous information. All of this stuff is documented.]

VMware does not yet support Opteron *Dual-Core* processors on any platform, but the DL585 Single-Core is a supported product.

[Malige]

I got back following information from VMware:

From the screen shot of the ESX crash, i see that the server was crashed because, the NMI (Non Maskable Interrupt) was occurred.

NMI occurs only when there is a faulty H/W on the system. I am still wondering how the HP H/W diagnostics and memtest came out clean without any errors.

[Paulo Meireles_1]

Hi,

We've been there, too. There seems to have been problems with the production of certain processor memory boards (a whole lot? who knows....) that need to be replaced, but HP is still looking into it.
We used memtest to check the machines, but memtest doesn't seem to be enough: machines that ran the test without a hitch for 10 days ended up crashing anyway while running VMware.
Memtest just runs on the first CPU, and is single-threaded. So, we thought, we'd ideally have a multithreaded test that ran on all CPUs. Our approach was to run memtest on 8 VMs, each 2 bound to a certain CPU, with 1.8GB of RAM each - so all RAM could be covered. This made some servers crash within minutes.
Anyway, this is a known issue within HP. I'm sure they'll come up with a solution.

[Mark Dees]

Hi,
I am curious if this problem with NMIs was ever dealt with by HP, or if it is an HP problem at all. I am experiencing a very similar issue on the DL-385 2P dual core running ESX 2.5.2. I am seeing NMIs in the VMware kernel logs produced when I put a load on the machine in only a single VM. The machine locks up requiring me ti use ILO to cycle the power. But none of the HP diagnostics or logs show any record of an NMI or any other error for that matter. I am running memtest86 without any errors. Is there any more information on this thread that someone might share?

[Andrew Mostovoy]

Just spend several hours chasing this problem. Latest info from HP is that there is a customer advisory on proc. boards with revision level lower than R. In one of my servers I had 2 rev AAs and 2 rev 0Ws. The second server had 3 0Ws and 1 AA. Replacements are on their way. Check those boards before running anything on these 585s.

[christoph_l]

Hi Andrew,
can you send a link to the customer advisory, because we have a similar problem. DL585 with 8 CPUs and 24GB of RAM. We are running ESX 3.0, on heavy load (filecopy or backup) the VM will die. ESX 3.0 is still running but we see NMI errors.

Thanx Christoph