SLES 12 SP1 OpenStack 6 fail Legacy Network Adapter PXE boot

Hi

I am trying to install SLES 12 SP1 with OpenStack 6 on a Hyper-V environment based on Windows 2012R2 latest updates.

In order to expand the OpenStack configuration with nodes a network or PXE boot is inevitable. In Hyper-V this feature is only available using the Legacy Network Adapter.

Doing so everything runs fine (though some installation steps are executed 3 or more times before they are finished) and the Suse OpenStack Console shows that the cloud-admin, control-1 and compute-1 are ok until you stop them, change for example the memory settings to more, and then the Suse VM start but does no longer connect to the OpenStack admin LAN network on the required Legacy Network Adapter.

The message given on the console is net eth0: 21140 transmit timed out, status fc… resetting… And this goes on forever.

Because it is an automatic installation of OpenStack nodes root access can only be done by the root user on the cloud-admin node. Since there is no connection nothing can be done on the VM node itself to resolve the issue.

Strangely enough I do not experience these issues on a Dell server but only on a HP Enterprise server, which is very strange because the OpenStack Admin LAN is a internal Hyper-V LAN, thus not connected to a physical network card (Software Only). Both the network and the network card are created by Microsoft and Suse and Microsoft claim full support.

Having nowadays these kind of issues surprises me because MS and Suse have been in a relationship for so long.

Does anyone knows what to do?

Thanks in advance
Roger

On 18/05/2016 13:54, rvunen1 wrote:
[color=blue]

I am trying to install SLES 12 SP1 with OpenStack 6 on a Hyper-V
environment based on Windows 2012R2 latest updates.

In order to expand the OpenStack configuration with nodes a network or
PXE boot is inevitable. In Hyper-V this feature is only available using
the Legacy Network Adapter.

Doing so everything runs fine (though some installation steps are
executed 3 or more times before they are finished) and the Suse
OpenStack Console shows that the cloud-admin, control-1 and compute-1
are ok until you stop them, change for example the memory settings to
more, and then the Suse VM start but does no longer connect to the
OpenStack admin LAN network on the required Legacy Network Adapter.

The message given on the console is net eth0: 21140 transmit timed out,
status fc… resetting… And this goes on forever.

Because it is an automatic installation of OpenStack nodes root access
can only be done by the root user on the cloud-admin node. Since there
is no connection nothing can be done on the VM node itself to resolve
the issue.

Strangely enough I do not experience these issues on a Dell server but
only on a HP Enterprise server, which is very strange because the
OpenStack Admin LAN is a internal Hyper-V LAN, thus not connected to a
physical network card (Software Only). Both the network and the network
card are created by Microsoft and Suse and Microsoft claim full support.[/color]

So with both Dell and HP servers set up exactly the same way you see
this issue with HP but not Dell?
[color=blue]

Having nowadays these kind of issues surprises me because MS and Suse
have been in a relationship for so long.

Does anyone knows what to do?[/color]

I don’t know Hyper-V but I do know SUSE OpenStack Cloud (which has it’s
own forum) so I’m working on possible differences between Dell and HP -
does the HP server have out-of-band/lights-out management enabled? If
so, could it be using the same IP subnet as for OpenStack Admin network
(192.168.124.0/24?

HTH.

Simon
SUSE Knowledge Partner

Yes that is correct. The HP server, a DL580 Gen7 has a standard Windows 2012R2 install with only the latest Emulex drivers installed (not the ones from HP). No HP software installed and HPsum never run.

The installation of Suse Cloud 6 is 100% the same: netwerk config mode dual, bastion network enabled, no changes to the proposed internal crowbar networks.

The network of the cloud-admin is eth0 Openstack Admin LAN (private hyper-V network) and eth1 Public LAN
The network of the nodes is: eth0 Openstack Admin LAN (private hyper-V network) and eth1 Openstack Trunk LAN (private hyper-V network).

On the nodes the eth0 adapter is a Legacy Network Adapter to enable network boot.

Both Microsoft and Suse claim full support of all Hyper-V features https://technet.microsoft.com/en-us/library/dn531027.aspx.

I had a similar experience with Windows Server 2016 preview 4 on the HP. We downgraded to exclude all issues due to the preview release.

Dell made a huge contribution to make sure that Suse OpenStack runs well on their systems. Check out /opt/dell and you find most of the stack there.

Roger

PS network problems are an issue in Suse on Hyper-V: https://www.novell.com/support/kb/doc.php?id=7008116