Bug 441158 - Libvirt loses track of domains that have been created and the user tries to create again
Summary: Libvirt loses track of domains that have been created and the user tries to c...
Keywords:
Status: CLOSED WORKSFORME
Alias: None
Product: Virtualization Tools
Classification: Community
Component: libvirt
Version: unspecified
Hardware: All
OS: Linux
low
medium
Target Milestone: ---
Assignee: Daniel Veillard
QA Contact: Martin Jenner
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2008-04-06 21:48 UTC by Henri Cook
Modified: 2010-03-16 17:13 UTC (History)
1 user (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2009-01-22 12:43:07 UTC
Embargoed:


Attachments (Terms of Use)
Config file for domain in question (788 bytes, text/xml)
2008-04-06 21:48 UTC, Henri Cook
no flags Details

Description Henri Cook 2008-04-06 21:48:58 UTC
Description of problem:

Occasionally 'virsh list' will say that a domain is not running when the process
is blatantly there in a ps faux - under libvirtd. This means the domain can't be
destroyed or created because of errors in both situations. Killing the process
results in a "[qemu-system-x86] <defunct>" process which can't be killed without
killing libvirtd and restarting it from init.d

Version-Release number of selected component (if applicable):

virsh -v == 0.4.0

How reproducible:

Very

Steps to Reproduce:
1. Create a domain
2. watch "virsh list" (in a seperate window)
3. Try to create the same domain again, the following error appears:
libvir: QEMU error : VM is already active
error: Failed to create domain from /home/virt/domains/Matilda/vm.cfg

Fair enough, but then the domain that was started previously disappears from
virsh list!
  
Actual results:

Please let me know if is reproduceible

Expected results:


Additional info:

Base System: 

Ubuntu Hardy latest over KVM

Comment 1 Henri Cook 2008-04-06 21:48:58 UTC
Created attachment 301440 [details]
Config file for domain in question

Comment 2 Daniel Veillard 2008-04-07 10:35:25 UTC
Changing Product, that's unrelated to Red Hat Enterprise or Fedora,
Please try with the latest upstream version i.e. libvirt-0.4.1 as 
many bugs were fixed since 0.4.0,

Daniel

Comment 3 Daniel Berrangé 2009-01-22 12:43:07 UTC
I've tried to reproduce this, but cannot. There should be no way for a defunct QEMU process not to be cleaned up by libvirt, because when we get a SIGHUP or EOF on the monitor socket we always call waitpid. If you can reproduce on latest upstream release (0.5.1) then please re-open.


Note You need to log in before you can comment on or make changes to this bug.