kvm instance fails to boot

302 views
Skip to first unread message

Attila Santo

unread,
Jan 31, 2011, 6:50:08 AM1/31/11
to gan...@googlegroups.com
hi,

i have installed ganeti 2.3.1 along kvm now on one node (debian squeeze rc2).

atm i'm struggling with following problem ->
when starting an instance i have created (gnt-instance add -t plain -n
node1.domain.com -o debootstrap+default --no-start --no-install -H
kvm:initrd_path=/boot/initrd-2.6-kvmU,cdrom_image_path=/home/os/debian/squeeze/squeeze_rc2_netinst.iso,boot_order=cdrom,vnc_bind_address=0.0.0.0
-B vcpus=2,memory=1024 -s 10G inst1.domain.com) ganeti reports the
instance as "running" with gnt-instance list

gnt-instance info inst1.domain.com reports:

Instance name: inst1.domain.com
UUID: 2597fd18-4478-4460-9998-e1ed25339d33
Serial number: 1
Creation time: 2011-01-31 12:09:18
Modification time: 2011-01-31 12:09:18
State: configured to be down, actual state is down
Nodes:
- primary: node1.domain.com
- secondaries:
Operating system: debootstrap+default
Allocated network port: 11002
Hypervisor: kvm
- console connection: vnc to node1.domain.com:11002 (display 5102)
- acpi: default (True)
- boot_order: cdrom
- cdrom_image_path: /home/os/debian/squeeze/squeeze_rc2_netinst.iso
- disk_cache: default (default)
- disk_type: default (paravirtual)
- initrd_path: /boot/initrd-2.6-kvmU
- kernel_args: default (ro)
- kernel_path: default (/boot/vmlinuz-2.6-kvmU)
- kvm_flag: default ()
- mem_path: default ()
- migration_downtime: default (30)
- nic_type: default (paravirtual)
- root_path: default (/dev/vda1)
- security_domain: default ()
- security_model: default (none)
- serial_console: default (True)
- usb_mouse: default ()
- use_chroot: default (False)
- use_localtime: default (False)
- vhost_net: default (False)
- vnc_bind_address: 0.0.0.0
- vnc_password_file: default ()
- vnc_tls: default (False)
- vnc_x509_path: default ()
- vnc_x509_verify: default (False)
Hardware:
- VCPUs: 2
- memory: 1024MiB
- NICs:
- nic/0: MAC: aa:00:00:d7:7f:7e, IP: 192.168.1.111, mode: routed, link:
Disks:
- disk/0: lvm, size 10.0G
access mode: rw
logical_id: xenvg/9aeeffa5-bed4-4221-8669-b4cb7e3a944a.disk0
on primary: /dev/xenvg/9aeeffa5-bed4-4221-8669-b4cb7e3a944a.disk0 (253:2)


gnt-cluster verify reports:

Mon Jan 31 12:36:40 2011 * Verifying global settings
Mon Jan 31 12:36:40 2011 * Gathering data (1 nodes)
Mon Jan 31 12:36:41 2011 * Gathering disk information (1 nodes)
Mon Jan 31 12:36:41 2011 * Verifying node status
Mon Jan 31 12:36:41 2011 * Verifying instance status
Mon Jan 31 12:36:41 2011 * Verifying orphan volumes
Mon Jan 31 12:36:41 2011 * Verifying orphan instances
Mon Jan 31 12:36:41 2011 * Verifying N+1 Memory redundancy
Mon Jan 31 12:36:41 2011 * Other Notes
Mon Jan 31 12:36:41 2011 - NOTICE: 1 non-redundant instance(s) found.
Mon Jan 31 12:36:41 2011 * Hooks Results


so to me all looks fine

but if i try to connect via vnc (knowing the console won't do it for
me because of kvm) i'm having following frozen screen

Starting SeaBIOS (version 0.5.1-20101112_115831-biber)
Decompressing Linux... Parsing ELF... done.
Booting the kernel.f to disable)... ok
_

its just frozen although ganeti tells me the instance ins running :puzzled:

at first i thought it was a kvm problem, but stopping ganeti and using
virt manager i can have a kvm instance up and running within a few
minutes.

whatever the problem is, i really hope its not something related to
strange permission settings (again) :D

i appreciate any help or hint i can get,

tia,

attila

Guido Trotter

unread,
Jan 31, 2011, 7:23:35 AM1/31/11
to gan...@googlegroups.com
On Mon, Jan 31, 2011 at 12:50 PM, Attila Santo <atti...@gmail.com> wrote:

Hi,

> so to me all looks fine
>
> but if i try to connect via vnc (knowing the console won't do it for
> me because of kvm) i'm having following frozen screen
>
> Starting SeaBIOS (version 0.5.1-20101112_115831-biber)
> Decompressing Linux... Parsing ELF... done.
> Booting the kernel.f to disable)... ok
> _
>
>
>
> its just frozen although ganeti tells me the instance ins running :puzzled:
>

This is normal. Ganeti cannot check "inside" the instance, it only
checks for the kvm process to be there.

> at first i thought it was a kvm problem, but stopping ganeti and using
> virt manager i can have a kvm instance up and running within a few
> minutes.
>

Well, probably it's related to the hypervisor parameters above. You
may have to specify the right ones for your kernel and environment
(boot disk, cd image, device types, and all this).
For example try changing kernel_path to None if you want to install
squeeze from cd.

> whatever the problem is, i really hope its not something related to
> strange permission settings (again) :D
>

Shouldn't be

thanks,

Guido

iBoris

unread,
Mar 4, 2011, 6:42:36 AM3/4/11
to ganeti
Hi,

I have the same frozen screen here:

Starting SeaBIOS (version 0.5.1-20101112_115831-biber)
Decompressing Linux... Parsing ELF... done.
Booting the kernel.f to disable)... ok

Have you solved this issue?

Thanks,
Boris

Attila Santo

unread,
Mar 4, 2011, 9:24:36 AM3/4/11
to gan...@googlegroups.com
hi iBoris,

yes:
make sure neither kernel_path nor initrd_path are set to anything. you
can check that by using
gnt-instance info YOURINSTANCE

if there is a value set try to modify those values with
gnt-instance modify -H kernel_path= YOURINSTANCE
gnt-instance modify -H initrd_path= YOURINSTANCE

of course you could also set those values cluster wide at the time you
initiate your cluster. i did so:

gnt-cluster init
--master-netdev eth0
--nic-parameters link=xen-br0
--enabled-hypervisors=kvm
-H kvm:
cdrom_image_path=/complete/path/to/your/desired/img.iso,
boot_order=cdrom,
vnc_bind_address=0.0.0.0,
kernel_path=,
initrd_path=
your.cluster.domain

hth,

Attila

2011/3/4 iBoris <bori...@gmail.com>:

iBoris

unread,
Mar 7, 2011, 6:05:35 AM3/7/11
to ganeti
Hi Attila,

Resetting kernel_path and initrd_path worked!

Thank you, very helpful,
Boris

On Mar 4, 2:24 pm, Attila Santo <attisa...@gmail.com> wrote:
> hi iBoris,
>
> yes:
> make sure neither kernel_path nor initrd_path are set to anything. you
> can check that by using
>         gnt-instance info YOURINSTANCE
>
> if there is a value set try to modify those values with
>         gnt-instance modify -H kernel_path= YOURINSTANCE
>         gnt-instance modify -H initrd_path= YOURINSTANCE
>
> of course you could also set those values cluster wide at the time you
> initiate your cluster. i did so:
>
> gnt-cluster init
>         --master-netdev eth0
>         --nic-parameters link=xen-br0
>         --enabled-hypervisors=kvm
>                 -H kvm:
>                         cdrom_image_path=/complete/path/to/your/desired/img.iso,
>                         boot_order=cdrom,
>                         vnc_bind_address=0.0.0.0,
>                         kernel_path=,
>                         initrd_path=
>         your.cluster.domain
>
> hth,
>
> Attila
>
> 2011/3/4 iBoris <borist...@gmail.com>:
Reply all
Reply to author
Forward
0 new messages