I've not yet tried 3.4, but I've noticed a problem with 3.2.1 and 3.3.0
where when a VM with a non-unique name is created, that similar symptoms
are observed. This was causing my VM restores from the SUSPENDED state
to fail, because ONE was having it go into a PROLOG instead of a
PROLOG_RESUME state (not remembering why at the moment, but something to
do with the value of the REASON or PxTIME values). I eventually was able
to figure out that the SQLPool object was getting confused, because I
think it assumes that every object that uses it must have a unique name.
This causes it to sometimes overwrite HISTORY values.
The attached patch to (to 3.3.0) solves my problem. I would have
submitted it upstream, but I might not consider it the "right" solution
and haven't had the time to get it into what I would consider a
reasonable state. But FWIW, here it is.
On Tue, Apr 17, 2012 at 09:53:07AM -0400, Carlos Martín Sánchez wrote:
> Hi Chris,
>
> It is indeed a bug. At first I wasn't able to reproduce the problem, then I
> realized you were creating VM directly from a template file rather than using
> onetemplate.
>
> We are looking into it [1]. Meanwhile, could you please try to create the VMs
> with "onetemplate instantiate" [2], and see it the problem disappears?
>
> Regards,
> Carlos
>
> [1] http://dev.opennebula.org/issues/1237
> [2] http://opennebula.org/documentation:rel3.4:vm_guide
>
> --
> Carlos Martín, MSc
> Project Engineer
> OpenNebula - The Open-source Solution for Data Center Virtualization
> www.OpenNebula.org | [email protected] | @OpenNebula
>
>
>
> On Mon, Apr 16, 2012 at 11:09 PM, Chris Johnston <[email protected]> wrote:
>
> Hi all,
>
> Has anyone else noticed that HISTORY records are corrupted when there are
> simultaneous VM actions? And some actions leave VMs, in particular a fail
> to livemigrate, with a bad history preventing a further live migration? It
> seems that update_history is failing at least for sqlite.
>
> # onevm create example.vm
>
> This results in a good history record...
>
> VIRTUAL MACHINE HISTORY
> SEQ HOSTNAME REASON START TIME PTIME
> 0 arch02 none 04/16 20:58:58 0d 00:00 0d 00:00
>
> <HISTORY_RECORDS>
> <HISTORY>
> <SEQ>0</SEQ>
> <HOSTNAME>arch02</HOSTNAME>
> <HID>3</HID>
> <STIME>1334609938</STIME>
> <ETIME>0</ETIME>
> <VMMMAD>vmm_kvm</VMMMAD>
> <VNMMAD>ovswitch</VNMMAD>
> <PSTIME>1334609938</PSTIME>
> <PETIME>1334609940</PETIME>
> <RSTIME>1334609940</RSTIME>
> <RETIME>0</RETIME>
> <ESTIME>0</ESTIME>
> <EETIME>0</EETIME>
> <REASON>0</REASON>
> </HISTORY>
> </HISTORY_RECORDS>
>
> # onevm create example.vm ; onevm create example.vm
>
> This results in none of the timers being set in the history record for
> either VM.
>
> VIRTUAL MACHINE HISTORY
> SEQ HOSTNAME REASON START TIME PTIME
> 0 host02 none - 15446d 20:5 0d 00:00
>
> <HISTORY_RECORDS>
> <HISTORY>
> <SEQ>0</SEQ>
> <HOSTNAME>host02</HOSTNAME>
> <HID>3</HID>
> <STIME>0</STIME>
> <ETIME>0</ETIME>
> <VMMMAD>vmm_kvm</VMMMAD>
> <VNMMAD>ovswitch</VNMMAD>
> <PSTIME>0</PSTIME>
> <PETIME>0</PETIME>
> <RSTIME>0</RSTIME>
> <RETIME>0</RETIME>
> <ESTIME>0</ESTIME>
> <EETIME>0</EETIME>
> <REASON>0</REASON>
> </HISTORY>
> </HISTORY_RECORDS>
>
> # onevm livemigrate 62 host2 ; onevm livemigrate 62 host01 (both of these
> fail on purpose)
> # onevm show 62
> <snip>
> STATE : ACTIVE
> LCM_STATE : RUNNING
> VIRTUAL MACHINE HISTORY
> SEQ HOSTNAME REASON START TIME PTIME
> 0 host02 user 04/16 20:58:58 0d 00:02 0d 00:00
> 1 host01 user - 15446d 21:0 0d 00:00
> 2 host02 user - 15446d 21:0 0d 00:00
> 3 host02 none 04/16 21:03:51 0d 00:00 0d 00:00
> 4 host02 none - 15446d 21:0 0d 00:00
> oneadmin@arch01:~$ onevm livemigrate 62 host02
> [VirtualMachineMigrate] Wrong state to perform action
>
> Chris
>
> ---------------------------------------------------------------------
> This transmission (including any attachments) may contain confidential
> information, privileged material (including material protected by the
> solicitor-client or other applicable privileges), or constitute non-public
> information. Any use of this information by anyone other than the intended
> recipient is prohibited. If you have received this transmission in error,
> please immediately reply to the sender and delete this information from
> your system. Use, dissemination, distribution, or reproduction of this
> transmission by unintended recipients is not authorized and may be
> unlawful.
> _______________________________________________
> Users mailing list
> [email protected]
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>
>
> _______________________________________________
> Users mailing list
> [email protected]
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
diff -urN opennebula-3.3.0.orig/src/pool/PoolSQL.cc opennebula-3.3.0/src/pool/PoolSQL.cc
--- opennebula-3.3.0.orig/src/pool/PoolSQL.cc 2012-02-20 06:31:34.000000000 -0500
+++ opennebula-3.3.0/src/pool/PoolSQL.cc 2012-03-10 02:44:18.000000000 -0500
@@ -21,6 +21,7 @@
#include <algorithm>
#include "PoolSQL.h"
+#include "VirtualMachinePool.h"
#include <errno.h>
@@ -218,8 +219,9 @@
string okey = key(objectsql->name,objectsql->uid);
name_index = name_pool.find(okey);
- if ( name_index != name_pool.end() )
- {
+ if (!dynamic_cast<VirtualMachinePool*>(this)) {
+ if ( name_index != name_pool.end() )
+ {
name_index->second->lock();
PoolObjectSQL * tmp_ptr = name_index->second;
@@ -228,10 +230,13 @@
pool.erase(tmp_ptr->oid);
delete tmp_ptr;
+ }
}
pool.insert(make_pair(objectsql->oid,objectsql));
- name_pool.insert(make_pair(okey, objectsql));
+ if (!dynamic_cast<VirtualMachinePool*>(this)) {
+ name_pool.insert(make_pair(okey, objectsql));
+ }
if ( olock == true )
{
@@ -401,7 +406,9 @@
string okey = key(tmp_ptr->name,tmp_ptr->uid);
pool.erase(index);
- name_pool.erase(okey);
+ if (!dynamic_cast<VirtualMachinePool*>(this)) {
+ name_pool.erase(okey);
+ }
delete tmp_ptr;
_______________________________________________
Users mailing list
[email protected]
http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
