[one-users] opennebula upgrade to 3.8.3 minor errors

[Daniel Molina]

On 6 February 2013 12:36, Andreas Calvo Gómez <andreas.calvo at scytl.com> wrote:
> Daniel,
>
> On 06/02/13 12:34, Daniel Molina wrote:
>>
>> On 6 February 2013 12:17, Andreas Calvo Gómez <andreas.calvo at scytl.com>
>> wrote:
>>>
>>> Hi Daniel!
>>>
>>> On 06/02/13 11:15, Daniel Molina wrote:
>>>>
>>>> Hi,
>>>>
>>>> On 1 February 2013 13:05, Andreas Calvo Gómez <andreas.calvo at scytl.com>
>>>> wrote:
>>>>>
>>>>> We have performed a update of both the OS (CentOS 6.2 to 6.3) and
>>>>> OpenNebula
>>>>> (3.8.1 to 3.8.3), and are facing some errors that weren't in the older
>>>>> release:
>>>>> - Logs can't be viewed from the webUI: selecting the "VM Log" tab shows
>>>>> an
>>>>> error message "Log for VM XXX not available".
>>>>
>>>> There was a bug in previous releases with the vm log directory, but it
>>>> should be fixed in 3.8.3. Did you restart the sunstone-server after
>>>> upgrading?
>>>
>>> Yes, several times because Sunstone or ONE becomes unresponsive from time
>>> to
>>> time.
>>
>> Is Sunstone running on the same machine as OpenNebula?
>
> Yes, we have a frontend with Sunstone and OpenNebula (using NFS to share the
> common storage).

The new vms logs should be in:
            system-wide: LOG_LOCATION + "/vms/#{id}.log"
            self-contained: LOG_LOCATION + "/vms/#{id}/vm.log"
the former directory was
            system-wide: LOG_LOCATION + "/#{id}.log"
            self-contained: LOG_LOCATION + "/#{id}/vm.log"

The logs from vms created in old versions have to be moved/linked to
this new directory. I will update the upgrade guide with this step,
sorry for the inconvenience..


>
>>
>>>>> - One node can't establish network connections: started VMs on one
>>>>> specific
>>>>> node don't have network connectivity, while they get the IP address by
>>>>> context. All other nodes work fine.
>>>>
>>>> Please check the following link:
>>>>
>>>>
>>>> http://wiki.opennebula.org/faq#my_vm_is_running_but_i_get_no_answer_from_pings_what_s_wrong
>>>>
>>>> Maybe there is something wrong with the host configuration.
>>>
>>> I'll try that link, but all nodes performed the same update procedure and
>>> only one is not working.
>>>
>>>>> - Sunstone/ONE seems to be stucked from time to time when performing
>>>>> large
>>>>> operations: as the old version (3.8.1), Sunstone becomes unresponsive
>>>>> and
>>>>> it
>>>>> is necessary to shut down all applications (one, sunstone, occi) and
>>>>> start
>>>>> them again. It usually happens when performing an operation with a
>>>>> large
>>>>> set
>>>>> of VMs.
>>>>>
>>>>> For all errors, no log in /var/log/one give any hint.
>>>>> Is there any other way to try to narrow down the root source of this
>>>>> problems?
>>>>
>>>> What are exactly the actions you performed and with how many vms?
>>>> Could you send us your sched.conf and oned.conf files?
>>>
>>> When a batch command is executed (usually deleting 50+ VMs), Sunstone or
>>> ONE
>>> becomes unresponsive and must be restarted.
>>> Yes, attached you will find our config files.
>>
>> Can you interact with OpenNebula through the cli, or is Sunstone the
>> only one that becomes unresponsive.
>
> Everything becomes unresponsive, but not at the same time.
> When a user performs a large batch operation, the same user can't run more
> commands, but other users are fine for a while.
> After some time, users can't log into Sunstone and other services (such as
> OCCI) become unresponsive too.
>

Could you send us the oned.log part when this operation is performed

Cheers

-- 
Daniel Molina
Project Engineer
OpenNebula - The Open Source Solution for Data Center Virtualization
www.OpenNebula.org | dmolina at opennebula.org | @OpenNebula