[Users] Problem running virtual machines

[jose garcia]

On 07/27/2012 11:51 AM, Itamar Heim wrote:
> On 07/27/2012 12:15 PM, jose garcia wrote:
>> On 07/27/2012 09:36 AM, Itamar Heim wrote:
>>> On 07/26/2012 07:38 PM, jose garcia wrote:
>>>> Apologies, version is 3.1.0-0.1.20120620git6ef9f8.fc17
>>>>
>>>> On 07/26/2012 05:18 PM, jose garcia wrote:
>>>>> Good evening,
>>>>>
>>>>> I am running oVirt 3.
>>>>>
>>>>> I am not sure if the source of the problem is oVirt or KVM/qemu but
>>>>> the machines run really slow since some days ago. There are some
>>>>> warnings regarding iscsi storage latency in the oVirt portal but the
>>>>> behaviour is also present with NFS storage.
>>>>>
>>>>> There are no visible errors either in engine.log or vdsm.log.
>>>>> Var/log/messages in the node shows over and over:
>>>>>
>>>>>     /usr/sbin/irqbalance: Load average increasing, re-enabling all
>>>>>     cpus for irq balancing
>>>>>
>>>
>>> what type of storage? what networking from host to storage?
>>
>> Good morning,
>>
>> I have got two nodes for an iSCSI datacenter and another for a NFS
>> datacenter. The machines run slow whatever the datacenter or node is
>> used (all run fedora 17, recently updated).
>>
>> Networking is basic. Only ovirtmgmt bridge and one network interface is
>> used per node.
>>
>> What is strange is that the VMs seem to run without any problem, only
>> they take more time to perform any activity. That, of course, causes
>> problems to ovirt-engine and the delays tend to set a node in a
>> non-operational state, specially when installing.
>
> VMs running slowly doesn't affect ovirt-engine.
> slow storage affects the SPM, which in turn affects ovirt-engine.
> what is the network bandwidth between nodes to storage? utilization? 
> how many spindles on the storage server? type of storage server?
>
Yes, you are right. Utilization and r_await are high.  I will perform 
some tests in order to determine what's changed. There are no network 
problems I am aware of. If It turns out to be something related to the 
OS, I will let you know.

Thank you.
>>
>>>
>>>
>>>>>
>>>>> and, after a while, waits in the node rise over 50%. It finally leads
>>>>> to a disconnection of the SPM node and a non-operational state.
>>>>>
>>>>> Has anyone any ideas why this happens?
>>>>>
>>>>> Regards,
>>>>>
>>>>> Jose Garcia
>>>>>
>>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Users mailing list
>>>> Users at ovirt.org
>>>> http://lists.ovirt.org/mailman/listinfo/users
>>>>
>>>
>>>
>>
>
>