natty-server-cloudimg-amd64 throws errors when using a ramdisk

[Juerg Haefliger]

I pulled natty-server-cloudimg-amd64.tar.gz, extracted the archive and then extracted the initrd from the image file. I then registered the three files (image, kernel, ramdisk) with Nova (2011.2-0ubuntu1) and launched an instance. The console log threw the following errors:

[ 0.451453] BIOS EDD facility v0.16 2004-Jun-25, 0 devices found
[ 0.452486] EDD information not available.
[ 0.603144] Freeing unused kernel memory: 880k freed
[ 0.605273] Write protecting the kernel read-only data: 10240k
[ 0.607776] Freeing unused kernel memory: 84k freed
[ 0.613642] Freeing unused kernel memory: 1412k freed
Loading, please wait...
Begin: Loading essential drivers ... done.
[ 0.628690] <30>udev[61]: starting version 167
Begin: Running /scripts/init-premount ... done.
Begin: Mounting root file system ... Begin: Running /scripts/local-top ... done.
Begin: Running /scripts/local-premount ... done.
[ 0.705857] 8139cp: 8139cp: 10/100 PCI Ethernet driver v1.3 (Mar 22, 2004)
[ 0.707364] ACPI: PCI Interrupt Link [LNKC] enabled at IRQ 11
[ 0.708441] 8139cp 0000:00:03.0: PCI INT A -> Link[LNKC] -> GSI 11 (level, high) -> IRQ 11
[ 0.712465] 8139cp 0000:00:03.0: eth0: RTL-8139C+ at 0xffffc90000312000, 02:16:3e:04:33:2e, IRQ 11
[ 0.714835] EXT4-fs (vda): mounted filesystem with ordered data mode. Opts: (null)
Begin: Running /scripts/local-bottom ... done.
done.
Begin: Running /scripts/init-bottom ... done.
udevd[61]: worker [104] did not accept message -1 (Connection refused), kill it

udevd[61]: worker [108] did not accept message -1 (Connection refused), kill it

udevd[61]: worker [109] did not accept message -1 (Connection refused), kill it

udevd[61]: worker [110] did not accept message -1 (Connection refused), kill it
udevd-work[183]: open /dev/null failed: No such file or directory

udevd-work[181]: open /dev/null failed: No such file or directory

Sometimes the instance would continue to successfully boot, sometimes it would hang at this point. The problem seems to be with the init-bottom script and specifically with the killing of udevd. Seems to be a timing issue since adding a simple sleep 1 before init-bottom is called 'fixes' the problem. Adding 'udevadm settle' in init-bottom right before the udev daemon is killed also seems to work. I suspect that the network events are lost due to the killed udev daemon and that somehow has a negative impact on the system.

Log with 'udevadm settle':

[ 0.312257] BIOS EDD facility v0.16 2004-Jun-25, 0 devices found
[ 0.312917] EDD information not available.
[ 0.461766] Freeing unused kernel memory: 880k freed
[ 0.462825] Write protecting the kernel read-only data: 10240k
[ 0.463993] Freeing unused kernel memory: 84k freed
[ 0.468530] Freeing unused kernel memory: 1412k freed
Loading, please wait...
Begin: Loading essential drivers ... done.
[ 0.489270] <30>udev[61]: starting version 167
Begin: Running /scripts/init-premount ... done.
Begin: Mounting root file system ... Begin: Running /scripts/local-top ... done.
Begin: Running /scripts/local-premount ... done.
[ 0.569248] EXT4-fs (vda): mounted filesystem with ordered data mode. Opts: (null)
Begin: Running /scripts/local-bottom ... [ 0.577919] 8139cp: 8139cp: 10/100 PCI Ethernet driver v1.3 (Mar 22, 2004)
[ 0.578770] ACPI: PCI Interrupt Link [LNKC] enabled at IRQ 11
[ 0.579415] 8139cp 0000:00:03.0: PCI INT A -> Link[LNKC] -> GSI 11 (level, high) -> IRQ 11
[ 0.582447] 8139cp 0000:00:03.0: eth0: RTL-8139C+ at 0xffffc90000312000, 02:16:3e:3b:6b:0b, IRQ 11
done.
done.
Begin: Running /scripts/init-bottom ... [ 0.633503] 8139too: 8139too Fast Ethernet driver 0.9.28
[ 0.644085] FDC 0 is a S82078B
done.

No errors and the instance boots up fine.

[Scott Moser (smoser)]

Could you please provide more information on "log with udev-settle" ?
What did you change?

[Scott Moser (smoser)]

sorry... i did not see your explanation of udev-settle in the bug description. Thanks for the complete bug report.

[Scott Moser (smoser)]

Other possibly related bugs are:
 * bug 712026 (note, fix-released, but we still saw apparent hang after that)
 * bug 712034 (duped to bug 647404, i'm not certain that was correct)
 * bug 647404 (possibly this is a dupe of that).