I have tried to xm save and restore a vm but xen or the kernel crash during
save; not always at the same point because the written save file is of
different size each time. Restore is of course not possible because the save
file is not complete.
I have tested this on a real P-III and inside a demo VMware 5.0 workstation.
I guess the problem could be with my kernel config, or with some libraries
of my distribution.
This is what the xend.log says (xend-debug.log and xfrd.log do not contain
anything usefull):
[2005-06-15 10:25:13 xend] INFO (XendMigrate:380) Save BEGIN: ['save',
['id', '1'], ['state', 'begin'], ['domain', '5'], ['file', '/dom1-save']]
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.save
['dom1', '5', 'begin', ['save', ['id', '1'], ['state', 'begin'], ['domain',
'5'], ['file', '/dom1-save']]]
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:487) domain_restart_schedule> 5
suspend 1
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.shutdown
['dom1', '5', 'suspend']
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:244) XendDomain>reap> domain
died name=dom1 id=5
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:247) XendDomain>reap> shutdown
id=5 reason=suspend
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.virq 4
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.suspended
['dom1', '5']
One other time doing xend restart crashed, too. Using vmware I was able to
see what happens at the console.
It said things like:
Failed to execute MMU updates.
invalid operand: 0000
CPU: 0
EIP: 0819:[<c016fe58>] Not tainted
EFLAGS: 00211286
eax: 00000022 ebx...
Process: python (pid: 3655, stackpage=c20a5000)<1>
Stack: c0248fa4 00009558 ...
Call Trace: ...
invalid operand: 0000
CPU: 0
EIP: 0819:[<c0105cfc>]
...
Call Trace: ...
The only interessing thing I can see in the call trace, is that the
addresses listed are repeating, i.e. it looks like an infinite loop.
Now the same for "xm save dom1 /dom1-save", it looks quite similar; EIP and
EFLAGS look the same. The call trace seems to be the same, again with
repeated address patterns. Process now says xfrd.
My setup:
Cent-OS 3.4
gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-49)
xen installed from xen-2.0.6-src.tgz (from the homepage)
kernel 2.4.30 (vanilla) + xen
default xenU kernel,
custom xen0 kernel (config attached)
I compiled the kernel by letting xen build process prepare it and then build
it with my own config (make mrproper, create config, make dep, etc.).
Drbd is used for storage (phy:drbd0).
/lib/tls is disabled.
Everthing else seems fine, I can create a vm, pause, unpause, destroy,
shutdown it. Networking from inside the vm works.
What can I do? What should I try? I am willing to help debugging, but I am
not skilled with gdb etc., so I would need help. I read some mails about
similar problems in the archives, but there did not seem to be any
conclusion.
Best Regards,
Michael Paesold
save; not always at the same point because the written save file is of
different size each time. Restore is of course not possible because the save
file is not complete.
I have tested this on a real P-III and inside a demo VMware 5.0 workstation.
I guess the problem could be with my kernel config, or with some libraries
of my distribution.
This is what the xend.log says (xend-debug.log and xfrd.log do not contain
anything usefull):
[2005-06-15 10:25:13 xend] INFO (XendMigrate:380) Save BEGIN: ['save',
['id', '1'], ['state', 'begin'], ['domain', '5'], ['file', '/dom1-save']]
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.save
['dom1', '5', 'begin', ['save', ['id', '1'], ['state', 'begin'], ['domain',
'5'], ['file', '/dom1-save']]]
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:487) domain_restart_schedule> 5
suspend 1
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.shutdown
['dom1', '5', 'suspend']
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:244) XendDomain>reap> domain
died name=dom1 id=5
[2005-06-15 10:25:13 xend] DEBUG (XendDomain:247) XendDomain>reap> shutdown
id=5 reason=suspend
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.virq 4
[2005-06-15 10:25:13 xend] INFO (XendRoot:112) EVENT> xend.domain.suspended
['dom1', '5']
One other time doing xend restart crashed, too. Using vmware I was able to
see what happens at the console.
It said things like:
Failed to execute MMU updates.
invalid operand: 0000
CPU: 0
EIP: 0819:[<c016fe58>] Not tainted
EFLAGS: 00211286
eax: 00000022 ebx...
Process: python (pid: 3655, stackpage=c20a5000)<1>
Stack: c0248fa4 00009558 ...
Call Trace: ...
invalid operand: 0000
CPU: 0
EIP: 0819:[<c0105cfc>]
...
Call Trace: ...
The only interessing thing I can see in the call trace, is that the
addresses listed are repeating, i.e. it looks like an infinite loop.
Now the same for "xm save dom1 /dom1-save", it looks quite similar; EIP and
EFLAGS look the same. The call trace seems to be the same, again with
repeated address patterns. Process now says xfrd.
My setup:
Cent-OS 3.4
gcc version 3.2.3 20030502 (Red Hat Linux 3.2.3-49)
xen installed from xen-2.0.6-src.tgz (from the homepage)
kernel 2.4.30 (vanilla) + xen
default xenU kernel,
custom xen0 kernel (config attached)
I compiled the kernel by letting xen build process prepare it and then build
it with my own config (make mrproper, create config, make dep, etc.).
Drbd is used for storage (phy:drbd0).
/lib/tls is disabled.
Everthing else seems fine, I can create a vm, pause, unpause, destroy,
shutdown it. Networking from inside the vm works.
What can I do? What should I try? I am willing to help debugging, but I am
not skilled with gdb etc., so I would need help. I read some mails about
similar problems in the archives, but there did not seem to be any
conclusion.
Best Regards,
Michael Paesold
Attached Files:
-
config (27.3 KB)