No network for RHEL VMware appliance

Dear lazyweb,

I installed a RHEL4 as a VMware guest in VMware Player, using the RHEL image from the VMware site. This RHEL instance has lots of trouble obtaining an IP address from its VMware host, resulting in DHCP request timeouts. I double, triple, and quad checked the settings on the guest, and on the host (the dhcpd part of the built-in vmware dhcp server), but without success. If I manually start dhclient, then I see the BOOTP/DHCP packets trying to leave the guest, but on the host I don't see anything entering the vmnet1 or vmnet8 interfaces. I tried the fix which suggests modifying /etc/sysconfig/networking-scripts/ifcfg-eth0, but without any success.

The strange thing is that I tried a Fedora guest also, which results in the same behaviour. When I install a Solaris 10 guest, the network works perfectly. This leads me to the conclusion that the problem resides on the Redhat side, and not on the VMware part.

Who can shed any light on this matter ?

Update : the solution (or at least a workaround).

Paul Cobbaut Fri, 04/11/2008 - 09:14

Hi,

i've been playing with RHEL4, RHEL5 and Fedora (and Solaris) in vmware (with ubuntu/debian/fedora as host)!

I never had a dhcp problem with the guests...some (obvious) suggestions:
- is the vmware setting set to bridged networking ?
- are the mac's unique ? (or did you duplicate the rhel4.vmx file ?) Try changing one digit in the last six bytes of the mac in the .vmx. or try removing the MACADDR or HWADDR lines in ifcfg-ethX.
- when cloning RHEL virtual machines, they often move to eth1 instead of eth0 (because of HWADDR)
- did you try removing ifcfg-ethX and/or using the RHEL GUI network tool to set it to dhcp ?
- are you using the gratis vmware server ?
- or just add a virtual nic in vmware and use that one.

anyway, it should work...it was never an issue for me (i use it for teaching, so we are talking about at least a hundred setups at ten different clients...)

kristof Fri, 04/11/2008 - 12:21

In reply to by Paul Cobbaut

Hi Paul,

first of all : I use VMware player, which means I cannot add disks, NICs or whatever.

While installing the Player, I enabled NAT and bridged networking, which results in the vmnet1 and vmnet8 interfaces. MACs in the guest are unique, and in the case of Fedora, bound to specific ifaces with udev rules. I tried the graphical tool eventually (when in despair), but that didn't solved anything (on the contrary...).
The weird thing is that both Fedora and RHEL result in the same behaviour, but Solaris works like a charm.

I would like to keep the VMware player setup, and not use the VMware Server, cause the purpose of the guest is to support a specific application, which will all run on a eeePC setup.

Kristof_ Sun, 04/13/2008 - 17:38

Hmmm. I tried a CentOS 5.1 image, which has no problems whatsoever in obtaining an IP address. I compared the setup of the network of both guests, made it on Fedora the same as on CentOS, I even copied the dhclient binary from CentOS to Fedora, but all without success...

dag Mon, 04/14/2008 - 07:21

I have seen very weird problems with software inside VMware and all of them were related to clock tick problems. Software often expects time to be linear, but in a VM time is not necessarily linear.

This can cause time to go faster inside a VM, but also to go slower in time. And if you have automatic time synchronisation enabled, a clock can be set backwords leading to very weird results.

There is good documentation about this from VMware.

Often changing the clock, or reducing the number of ticks per seconds to 250 or 100 can reduce the number of lost ticks and as a result "fix" software that has time-related issues. What I normally do to see if my
system is affected, is running a dstat like:

# dstat -t -i -I0

This will monitoring the number of time interrupts per seconds. This way you can see how many ticks per seconds your system has (often referenced as HZ) and how much VMware is generating for the VM.

Not sure if your dhcp problem is related to this, but it may be worth a try.

Kristof_ Mon, 04/14/2008 - 07:26

In reply to by dag

I doubt a clock tick issue might be the cause; in the CentOS image, there's a clear clock tick issue (time goes really fast there - probably enlarged tickiness to enhance the responsiveness of the guest), and that one works. Or I might to increase the number of ticks in the RHEL image... Will give it a try.

Anonymous Fri, 03/20/2009 - 16:13

Bit late, but have you tried removing the MAC address from the fedora network configuration. Seems like VMware changes the MAC every time, and if you have it set in the config (probably assigned during install) but it's different, then fedora networking will just "not work".