[Xen-users] Live migration failures

[Igor Chubin <igor@xxxxxxx>]

Hello, 

I have a problem with live migration of Xen domains
(or not a problem may be, but a behaviour that I can't explain).

I have two hosts running as Xen domains 0.
There are several Xen U-domains running at these hosts.

I have remarked weird error during live migration of my U-domains.

Live migration of domain fails if the domain is migrating for 
the first time (Figure 1).

If the domain have been ever executed on the target host 
before the migration, 
all works fine (Figure 2).
The migration in the reverse direction (from host2 to host1)
and all of the domain migration that follows afterwards
accomplishes without any problems.



         HOST1                            HOST2


|                             |                              | 
| create dom1                 |                              | 
| dom1 is running             |                              | 
| migrate dom1                |                              | 
|                        ------->X                           | 
|                             |(fail)                        | 
|                             |                              | 
|                             |                              | 

----------------------------------------------------------------
Figure 1.




         HOST1                            HOST2


|                             |    create dom1               | 
|                             |    dom1 is running           | 
|                             |    destroy dom1              | 
| create dom1                 |                              | 
| dom1 is running             |                              | 
| migrate dom1                |                              | 
|                        ------->                            | 
|                             |(ok)                          | 
|                             |                              | 
|                             |    dom1 is running           | 
|                             |    migrate dom1 back         | 
|                          <-------                          | 
|                        (ok) |                              | 
|                             |                              | 
| dom1 is running             |                              | 
|                             |                              | 

----------------------------------------------------------------
Figure 2.



I am aware of the similar discussion:
* http://lists.xensource.com/archives/html/xen-devel/2007-01/msg00249.html

I know that in that case the problem was solved by turning PAE off.

Is there other ways to solve the problem?


Thank you!



My Xen hypervisor version:
=====
$ SUDo xm dmesg | head
Xen version 3.0.3-1 (Debian 3.0.3-0-2) (ultrotter@xxxxxxxxxx) (gcc version 
4.1.2 20061028 (prerelease) (Debian 4.1.1-19)) Thu Nov 2 20:32:28 CET 2006
Latest ChangeSet: Tue Oct 17 22:09:52 2006 +0100 

(XEN) Command line: /boot/xen-3.0.3-1-i386-pae.gz dom0_mem=192M
=====


The part of xend log file, 
that concerns experinced problem:
(key line is marked with >>> <<< symbols)

=====
[2007-04-04 22:05:49 xend 3280] DEBUG (__init__:1072) Balloon: 1873316 KiB 
free; need 73728; done.
[2007-04-04 22:05:49 xend 3280] DEBUG (__init__:1072) [xc_restore]: 
/usr/lib/xen-3.0.3-1/bin/xc_restore 18 5 18432 1 2
[2007-04-04 22:05:49 xend 3280] ERROR (__init__:1072) xc_linux_restore start: 
max_pfn = 4800
[2007-04-04 22:05:49 xend 3280] ERROR (__init__:1072) Increased domain 
reservation by 12000 KB
[2007-04-04 22:05:49 xend 3280] ERROR (__init__:1072) Reloading memory pages: 0%
[2007-04-04 22:05:50 xend 3280] ERROR (__init__:1072) Received all pages (0 
races)
>>> [2007-04-04 22:05:50 xend 3280] ERROR (__init__:1072) Failed to pin batch 
>>> of 15 page tables: 22 <<<
[2007-04-04 22:05:50 xend 3280] ERROR (__init__:1072) Restore exit with rc=1
[2007-04-04 22:05:50 xend.XendDomainInfo 3280] DEBUG (__init__:1072) 
XendDomainInfo.destroy: domid=5
[2007-04-04 22:05:50 xend.XendDomainInfo 3280] DEBUG (__init__:1072) 
XendDomainInfo.destroyDomain(5)
[2007-04-04 22:05:50 xend.XendDomainInfo 3280] ERROR (__init__:1072) 
XendDomainInfo.destroy: xc.domain_destroy failed.
=====


-- 
WBR, i.m.chubin


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users