Connection error when use PXE+NFS to boot guest

[Clotho Tsang]

I use NFS+PXE to boot RHEL 5. The installation is slow, and I send Ctrl+Alt+F4, then find that half (but not all) of connections are dropped:

nfs: server 192.168.122.50 OK
nfs: server 192.168.122.50 OK
nfs: server 192.168.122.50 OK
nfs: server 192.168.122.50 not responding, still trying
nfs: server 192.168.122.50 OK
nfs: server 192.168.122.50 not responding, still trying
nfs: server 192.168.122.50 not responding, still trying
:
:

I have checked that there is no IP conflict.

Host machine
Kernel: 2.6.32.9-70.fc12.x86_64
QEMU version: 0.12.3-2.fc12.x86_64
libvirt version: 0.7.1-15.fc12.x86_64

Guest machine 1 (NFS server)
Kernel: 2.6.18-164.el5
nfs-utils version: 1.0.9-42.el5
Set up command: virt-install --connect qemu:///system -n m5 --vnc --accelerate --ram=256 --noreboot --os-type=linux --pxe --os-variant=rhel5 --mac=52:54:00:50:00:00 --file=/dev/VolGroup00/m5

Guest machine 2 (the machine to be installed via PXE+NFS)
Set up command: virt-install --connect qemu:///system -n e51 --vnc --accelerate --ram=128 --noreboot --os-type=linux --pxe –os-variant=rhel5 --mac=52:54:00:51:01:00 --file=/dev/VolGroup00/e51

============================================================================
I try to use NFS under different situations to test.
Test 1) Try normal NFS mount. This test successes.
Test 2) Try to boot a diskless machine via PXE, use a NFS mount point at root. This test suffers the same problem

At test 1)
NFS client mount an export of the NFS server, and use command "find ." to see if it runs smoothly.
Result: it runs smoothly for thousands of lines.

Guest machine 3 (NFS server)
Kernel: 2.6.18-164.el5
Set up command: virt-install --connect qemu:///system -n dm1 --vnc --accelerate --ram=256 --noreboot --os-type=linux --pxe --os-variant=rhel5 --mac=52:54:00:d1:00:00 --file=/dev/VolGroup00/dm1

NFS client is Guest machine 1 mentioned before.

At test 2)
Set up a PXE server, write a customized initrd, in which init script uses "mkrootdev -t nfs -o nolock,ro" to set root directory on a NFS mount.

Result: the client machine can boot, but very slowly, with numerous
nfs: server 192.168.122.110 OK
nfs: server 192.168.122.110 not responding, still trying

NFS server: it is Guest machine 3
Client to be booted: Guest machine 4
Set up command: virt-install --connect qemu:///system -n de11 --vnc --accelerate --ram=128 --noreboot --os-type=linux --pxe –os-variant=rhel5 --nodisk --mac=52:54:00:d1:01:00

[Clotho Tsang (clothohk)]

I have also tried to use qemu 0.11.0 and 0.10-16, the bugs are the same.

I have seen similar bug report in the following. Although they said the bug is fixed in version >= 0.11.1, but I still find them at 0.12
http://<email address hidden>/msg17212.html
http://<email address hidden>/msg17357.html

[Thomas Huth (th-huth)]

Triaging old bug tickets ... can you still reproduce this issue with the
latest version of QEMU (currently version 2.8)?

[Launchpad Janitor (janitor)]

[Expired for QEMU because there has been no activity for 60 days.]