SLES 11 for SBS-Evaluation gives PSOD on ESXi 4.1-Host

SUSE Linux Enterprise Server SLES Install-Boot
1

The esxi 4.1-host ist on build 1682698 and holds some virtual machines with Netware (6.0 SBS), Windows (Server 2008R2) and Linux (Fedora 20).

For evaluation I downloaded the “Evaluation Novell Open Workgroup Suite for Small Business (6 files) dated 12 Jul 2013”.

During the normal install as “fully-virtualized” with the graphical-gui the install freezes at some point bringing the esxi-host in a non-responsive mode. For that reason I started the install with the kernel-option “textmode=1” where the install ended in blowing the esxi-host with a PSOD and the message “recursive panic on same cpu”.

That´s impressive! :wink:

In my poor eyes it should not be possible to blow the host in that way. Because I had sort of the same results with OpenSuse (but not Fedora) question came up, if there are any specific settings or steps for the vm to be taken, which I could not find myself so far.

The vm-setup includes vm-version 7, guest as Linux SLES11-64bit, 1 processor, 2 GB memory, 1 Nic E1000 with manual mac-address for the DNS-Server, SCSI LSI Logic Parallel, new disk with 12 GB as permanent, graphics with 4 MB memory.

Any ideas?

Sincerely

Karl

2

On 01/01/2015 10:34 AM, Karl Kunze wrote:[color=blue]

The esxi 4.1-host ist on build 1682698 and holds some virtual machines
with Netware (6.0 SBS), Windows (Server 2008R2) and Linux (Fedora 20).

For evaluation I downloaded the “Evaluation Novell Open Workgroup Suite
for Small Business (6 files) dated 12 Jul 2013”.

During the normal install as “fully-virtualized” with the graphical-gui
the install freezes at some point bringing the esxi-host in a
non-responsive mode. For that reason I started the install with the
kernel-option “textmode=1” where the install ended in blowing the
esxi-host with a PSOD and the message “recursive panic on same cpu”.

That�s impressive! :wink:

In my poor eyes it should not be possible to blow the host in that way.
Because I had sort of the same results with OpenSuse (but not Fedora)
question came up, if there are any specific settings or steps for the vm
to be taken, which I could not find myself so far.

The vm-setup includes vm-version 7, guest as Linux SLES11-64bit, 1
processor, 2 GB memory, 1 Nic E1000 with manual mac-address for the
DNS-Server, SCSI LSI Logic Parallel, new disk with 12 GB as permanent,
graphics with 4 MB memory.[/color]

Is the ESXi platform a VMware supported platform? The reason I ask is that the
only time I ever got an unsuspected PSOD was when running on an unsupported
platform. For whatever reason, and this is especially true of version 4, VMware
(beyond 4.0, it was ok) decided to do some very specific things. For example if
your ESXi host has two socketed CPUs and they aren’t perfectly matched (and I
mean perfectly)… that can case the PSOD.

Not saying there isn’t something else going on, just saying that it put a very
bad taste in my mouth for VMware. You won’t have this problem with Xen or KVM btw.

VMware (esp. v4.1) is designed for very specific platforms only… you have been
warned.

3

[QUOTE=cjcox;25606]On 01/01/2015 10:34 AM, Karl Kunze wrote:[color=blue]

The esxi 4.1-host ist on build 1682698 and holds some virtual machines
with Netware (6.0 SBS), Windows (Server 2008R2) and Linux (Fedora 20).

For evaluation I downloaded the “Evaluation Novell Open Workgroup Suite
for Small Business (6 files) dated 12 Jul 2013”.

During the normal install as “fully-virtualized” with the graphical-gui
the install freezes at some point bringing the esxi-host in a
non-responsive mode. For that reason I started the install with the
kernel-option “textmode=1” where the install ended in blowing the
esxi-host with a PSOD and the message “recursive panic on same cpu”.

That�s impressive! :wink:

In my poor eyes it should not be possible to blow the host in that way.
Because I had sort of the same results with OpenSuse (but not Fedora)
question came up, if there are any specific settings or steps for the vm
to be taken, which I could not find myself so far.

The vm-setup includes vm-version 7, guest as Linux SLES11-64bit, 1
processor, 2 GB memory, 1 Nic E1000 with manual mac-address for the
DNS-Server, SCSI LSI Logic Parallel, new disk with 12 GB as permanent,
graphics with 4 MB memory.[/color]

Is the ESXi platform a VMware supported platform? The reason I ask is that the
only time I ever got an unsuspected PSOD was when running on an unsupported
platform. For whatever reason, and this is especially true of version 4, VMware
(beyond 4.0, it was ok) decided to do some very specific things. For example if
your ESXi host has two socketed CPUs and they aren’t perfectly matched (and I
mean perfectly)… that can case the PSOD.

Not saying there isn’t something else going on, just saying that it put a very
bad taste in my mouth for VMware. You won’t have this problem with Xen or KVM btw.

VMware (esp. v4.1) is designed for very specific platforms only… you have been
warned.[/QUOTE]

It is not a VMware-certified system, but it ran stable (with Netware etc.) since years. We are in the test-lab here, the production-system itself is VMware-certified. Both are single-socketed-cpu-machines. After seeing this PSOD I am unsure, whether I should throw this evaluation on the production-system, just to see, if your guess ist right.

On the other hand these hangs/PSODs are reproducable for me with Suse-products only so far. My hope is, that there is something wrong with my setting and that there are tricks, to get beyond this trouble somehow.

Sincerely

Karl

4

On 01/02/2015 03:44 AM, Karl Kunze wrote:
…snip[color=blue]

It is not a VMware-certified system, but it ran stable (with Netware
etc.) since years. We are in the test-lab here, the production-system
itself is VMware-certified. Both are single-socketed-cpu-machines. After
seeing this PSOD I am unsure, whether I should throw this evaluation on
the production-system, just to see, if your guess ist right.

On the other hand these hangs/PSODs are reproducable for me with
Suse-products only so far. My hope is, that there is something wrong
with my setting and that there are tricks, to get beyond this trouble
somehow.[/color]

My case was repeatable as well. You might want to try ESXi 5.x etc… if
possible. I have a feeling that it tolerates hardware better. Not sure why I
say that though.

My solution at the time was move to Xen, but on that same system, I’m now
running KVM. I’ll continue to work with VMware, but only on certified hardware.

5

[QUOTE=cjcox;25615]On 01/02/2015 03:44 AM, Karl Kunze wrote:
…snip[color=blue]

It is not a VMware-certified system, but it ran stable (with Netware
etc.) since years. We are in the test-lab here, the production-system
itself is VMware-certified. Both are single-socketed-cpu-machines. After
seeing this PSOD I am unsure, whether I should throw this evaluation on
the production-system, just to see, if your guess ist right.

On the other hand these hangs/PSODs are reproducable for me with
Suse-products only so far. My hope is, that there is something wrong
with my setting and that there are tricks, to get beyond this trouble
somehow.[/color]

My case was repeatable as well. You might want to try ESXi 5.x etc… if
possible. I have a feeling that it tolerates hardware better. Not sure why I
say that though.

My solution at the time was move to Xen, but on that same system, I’m now
running KVM. I’ll continue to work with VMware, but only on certified hardware.[/QUOTE]

Hmmm, ok. Still wondering, why the underlying hardware should be a reason for the host to be banged by the guest. To get rid of that sort of trouble was one reason for virtualization by hypervisor, wasn´t it?.

I guess moving the host towards ESXi 5.x should be my next step. I am unsure, what happens with the Netware-VMs, though. There seems to be no support for that with ESXi 5.x, but I will do some further research on that.

Thank you very much for your input.

Sincerely

Karl

6

I didnt find how to start my own thread… so i will comment on yours.
I bought suse 9.1 , in a box set at comp usa in 2007… or was it 8 ? i bought it because at the time , i was still on dial up.
not to mention the fact that having the boxset and all the documentation in original form was far more prestigious.
now everyone can download suse … and I believe that the suse downloads should be offered in torrent form as well as conventional dowloads.
Regular downloads are prone to failure due to browser hangups. pauses and other unrelated issues.
Not to mention that torrents are far faster due to the exponential bandweith growth.
If you can point to where i should be directing my comment , feel free to write to Marden… I am at : EEvergreentree at aol .com
Thank you
( suse user since 9.1)

7

On 01/03/2015 05:44 AM, Karl Kunze wrote:[color=blue]

cjcox;25615 Wrote:[color=green]

On 01/02/2015 03:44 AM, Karl Kunze wrote:
…snip[color=darkred]

It is not a VMware-certified system, but it ran stable (with Netware
etc.) since years. We are in the test-lab here, the production-system
itself is VMware-certified. Both are single-socketed-cpu-machines.[/color]
After[color=darkred]
seeing this PSOD I am unsure, whether I should throw this evaluation[/color]
on[color=darkred]
the production-system, just to see, if your guess ist right.

On the other hand these hangs/PSODs are reproducable for me with
Suse-products only so far. My hope is, that there is something wrong
with my setting and that there are tricks, to get beyond this trouble
somehow.[/color]

My case was repeatable as well. You might want to try ESXi 5.x etc…
if
possible. I have a feeling that it tolerates hardware better. Not sure
why I
say that though.

My solution at the time was move to Xen, but on that same system, I’m
now
running KVM. I’ll continue to work with VMware, but only on certified
hardware.[/color]

Hmmm, ok. Still wondering, why the underlying hardware should be a
reason for the host to be banged by the guest. To get rid of that sort
of trouble was one reason for virtualization by hypervisor, wasn�t it?.[/color]

It’s weird, but for whatever reason, 4.1 (I can’t remember if it was after a
patch to 4.0, or when exactly)… for whatever reason, VMware started “checking”
for things that in my opinion don’t matter, and then it would throw a PSOD if it
didn’t like what it saw. The messages didn’t really indicate that was the
reason, I found that out through my own research into the problem.

VMware has always been sort of “evil”… but there was period when they played
around with being “very evil”… I think they are less evil now esp. since the
number one cloud provider doesn’t use them.

Off topic, but VMware promised browser based mgmt and delivered (v5) a solution
totally built on Flash… (can we say “evil” or “stupid”?). VMware and and it’s
kludge (and it’s a big kludge) known as VCenter need to die. I mean, they’ve
got money, but don’t really seem to care about delivering a quality product even
in the midst of ever rising competition (confused?).

Sure… there are things that VMware does “ok” that others don’t have, but
VMware doesn’t seem interested in making “ok” into “good”… so eventually,
they’ll get surpassed by the competition (my prediction).
[color=blue]

I guess moving the host towards ESXi 5.x should be my next step. I am
unsure, what happens with the Netware-VMs, though. There seems to be no
support for that with ESXi 5.x, but I will do some further research on
that.[/color]

You might want to check our alternatives to VMware (e.g. Xen, I think you can
even use Xen that comes with SLES). Eventually KVM will rule… but none of the
choices (openStack, oVirt/RHEV, etc.) are awesome… but they are getting there

8

I am not thinking in categories like good and evil. I am seeking for a working solution and from that point of view VMware ESXi 4.1 does its job.

While continuing with tests I found out, that installing SLES with NIC as VMXNET3 instead of E1000 works perfect. The server is up an running since some days now, so I can move on with my tests.

Just in case someone else runs in that problem.

Sincerely

Karl

9

On 01/06/2015 11:54 AM, Karl Kunze wrote:[color=blue]

I am not thinking in categories like good and evil. I am seeking for a
working solution and from that point of view VMware ESXi 4.1 does its
job.

While continuing with tests I found out, that installing SLES with NIC
as VMXNET3 instead of E1000 works perfect. The server is up an running
since some days now, so I can move on with my tests.

Just in case someone else runs in that problem.
[/color]

And one advantage of VMware over KVM virtio and Xen is that VMXNET3.

It’s the only virtual NIC I’ve been able to drive to 10Gbit and beyond.

Hope things go well with Netware and the rest.