Bug 994882 - Live block migration results in a zero console log
Summary: Live block migration results in a zero console log
Keywords:
Status: CLOSED NOTABUG
Alias: None
Product: Virtualization Tools
Classification: Community
Component: libvirt
Version: unspecified
Hardware: x86_64
OS: Linux
unspecified
high
Target Milestone: ---
Assignee: Libvirt Maintainers
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2013-08-08 07:58 UTC by quqi99
Modified: 2016-02-25 14:11 UTC (History)
5 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2016-02-25 14:11:59 UTC
Embargoed:


Attachments (Terms of Use)

Description quqi99 2013-08-08 07:58:40 UTC
Description of problem:
live-migration with --block-migrate results in a zero console log. 

Version-Release number of selected component (if applicable):
libvirt version: 0.10.2, package: 18.el6

How reproducible:
always


Steps to Reproduce:
1. configure share ssh pubkey between the two hosts.
2. do block live mgiration by libvirt's CLI:
   virsh migrate --live --copy-storage-all --verbose <vm-name> qemu+ssh://<target-host>/system
   or by virt-manager refer in the link:
   https://access.redhat.com/site/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Virtualization_Administration_Guide/sect-Virtualization-KVM_live_migration-Migrating_with_virt_manager.html
3. results in a zero console log

Actual results:
can migrate successfully, but there is a zero console.log in target host.


Expected results:
hope the console.log in source host and target host are the same.


Additional info:
The console log is necessary to diagnose many kinds of problems, including errors reported by Anaconda, problems with the boot images, a misconfigured VM, failures in the harness, network problems communicating with Beaker...

Comment 1 Kashyap Chamarthy 2015-05-11 13:10:31 UTC
Just trying to understand your question, why would you expect the console log (which is related to boot) be updated when live migration is performed?

Comment 2 Pawel Koniszewski 2015-08-10 12:06:58 UTC
Kashyap, not updated but moved to destination host along with the VM.

Please consider situation when operator live migrates VM only for host maintenance. After host maintenance he will live migrate VM back to original host and I think that in such case console.log should still be there (e.g. transferred to destination host with VM and then back to original host).

Comment 3 Kashyap Chamarthy 2016-02-25 14:01:15 UTC
(In reply to Pawel Koniszewski from comment #2)
> Kashyap, not updated but moved to destination host along with the VM.
> 
> Please consider situation when operator live migrates VM only for host
> maintenance. After host maintenance he will live migrate VM back to original
> host and I think that in such case console.log should still be there (e.g.
> transferred to destination host with VM and then back to original host).


Pawel, seems like there's general agreement from libvirt upstream that this needs to be handled at a higher-layer than libvirt.

[Below is a comment from IRC, by Dan Berrangé, on this topic]

"This bug is basically asking for the block migration functionality to extend to other qemu devices backed by files.  This is pretty tricky in general, because with latest libvirt & qemu, qemu doesn't even get acess to the log files.  It just gets given an anonymous pipe file descriptor, so it has no ability to migrate the log file even if it wanted to.

libvirt tries to only concern itself with what happens on a single-node 
so moving stuff between nodes is left to the app using libvirt which knows better what to do".

Comment 4 Kashyap Chamarthy 2016-02-25 14:11:59 UTC
Pawel agrees (on #openstack-nova on Freenode) that, it "sounds fair that higher-layer should handle this, yes"

So, closing the bug with the rationale in comment#3


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