[Xen-devel] Migration failed due to "No extended context"

[Kieran Mansley <kmansley@xxxxxxxxxxxxxx>]

We have seen migration occasionally failing (occasionally meaning there
have been hundreds of successful migrations before the failure) on an
automated test that just cycles virtual machines amongst a set of
physical servers in the presence of network traffic to/from the
guests.  

/var/log/xen/xend.log shows this:

[2008-02-07 13:30:40 5149] INFO (XendCheckpoint:365) ERROR Internal
error: No
extended context for VCPU0
[2008-02-07 13:30:40 5149] INFO (XendCheckpoint:365) Save exit rc=1
[2008-02-07 13:30:40 5149] ERROR (XendCheckpoint:141) Save failed on
domain
lynx01-v1 (80).
Traceback (most recent call last):
  File
"xen-
unstable.hg/dist/install/usr/lib/python/xen/xend/XendCheckpoint.py",
line 109, in save
    forkHelper(cmd, fd, saveInputHandler, False)
  File
"xen-
unstable.hg/dist/install/usr/lib/python/xen/xend/XendCheckpoint.py",
line 353, in forkHelper
    raise XendError("%s failed" % string.join(cmd))
XendError: /usr/lib/xen/bin/xc_save 24 80 0 0 1 failed

The error seems to stem from tools/libxc/xc_domain_save.c, which in turn
is due to an error in xen/arch/x86/domctl.c's handling of
XEN_DOMCTL_get_ext_vcpucontext, although I have tracked down which bit
of that is returning an error.

There is a corresponding error when the domain tries to resume after the
failed suspend, again reporting that there is no extended context.

I wondered if it could be related to this:

http://lists.xensource.com/archives/html/xen-devel/2007-11/msg00814.html

The patch supplied in that report doesn't seem to have been applied.

This is based on a recent checkout of xen-unstable.hg

Kieran



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