Re: [ovirt-users] PCI Passthrough problem

2016-06-08 Thread Jérôme Chapelle
Thanks to all of us ! 
We read the Sandro's email annoucement: 
http://lists.ovirt.org/pipermail/users/2016-June/040172.html 
And applied to the HV the update process pointed in: 
http://www.ovirt.org/release/3.6.7/ 
(we only updated vdsm) 
Now our HV is working fine and we are able again to do pci passthrough with it. 
As a result, the virtual machine hosted on it now sees tape. 
Thanks again for your fast and precisous help ! 
Jerome 


From: "Michal Skrivanek"  
To: "Lionel Caignec"  
Cc: "honvault" , "Martin Polednik" , 
"users"  
Sent: Wednesday, June 8, 2016 9:12:44 AM 
Subject: Re: [ovirt-users] PCI Passthrough problem 

> On 08 Jun 2016, at 08:41, Lionel Caignec  wrote: 
> 
> Hi thanks for reply 
> 
> There is something i does not understand, on my hosts i've vdsm 4.17.28-0.el7 
> no indication about a version 3.6.5 or 3.6.7. 
> How can i get this version? 

The ovirt release 3.6.7 rc2, see Sandro’s announcement email sent to this list 
last Thursday. Once you reconfigure repos as explained there you will see the 
new fixed vdsm:) It’s a release candidate repo 

Hope it works, 
michal 

> 
> 
> 
> 
> - Mail original - 
> De: "Martin Polednik"  
> À: "Lionel Caignec"  
> Cc: "Martin Polednik" , "chapelle" , 
> "users" , "honvault"  
> Envoyé: Mercredi 8 Juin 2016 08:20:02 
> Objet: Re: PCI Passthrough problem 
> 
> On 07/06/16 19:04 +0200, Lionel Caignec wrote: 
>> Hi i'm coworker of Mr Chapelle, 
>> 
>> Thank you for your reply. 
>> 
>> so if i understand we need to "downgrade" our ovirt engine which is in 
>> version 3.6.6.x to version 3.6.5? 
>> But how can we do that? 
> 
> You would only need to downgrade the 2 hosts to 3.6.5 VDSM. But 
> thinking of it, you could instead upgrade to 3.6.7 (rc2) - the bug is 
> fixed there. 
> 
>> Lionel Caignec. 
>> 
>> 
>> 
>> - Mail original - 
>> De: "Martin Polednik"  
>> À: "chapelle"  
>> Cc: "users" , "honvault"  
>> Envoyé: Mardi 7 Juin 2016 17:11:10 
>> Objet: Re: [ovirt-users] PCI Passthrough problem 
>> 
>> On 07/06/16 16:50 +0200, Jérôme Chapelle wrote: 
>>> We are no more able to give a pci device to one VM. 
>>> Here is the process we used to assign one device to one VM: 
>>> - Click on the VM, and then on the tab "Host devices" 
>>> - In the tab, click on "pin to host" and select the host that will give the 
>>> pci device 
>>> - Click on "add device", in the window appear all the pci devices: choose 
>>> the pci card you wish (ex. pci__83_00_0 pci__83_00_1) and then 
>>> click on the gray arrow to assign this device to the VM 
>>> - Click on ok 
>>> - boot the VM. 
>>> The problem is that in the window that appears, there is not a single pci 
>>> device listed (there should be many listed). 
>>> I checked first: the IOMMU setting is set in the kernel. 
>>> 
>>> Our environment: 
>>> We have two HV running CentOs7. 
>>> Each of them have three fiber channel cards. First card is connected to a 
>>> disk array, second one to the first tape library and third one to another 
>>> tape library. 
>>> We used to give the second card to a VM1, and the third card to another 
>>> VM2. Both VMs ran on one HV or on the other in the past: everything was 
>>> fine. 
>>> What happened: 
>>> Today we add to upgrade one of the HV (HV1). Both VMs ran on this HV1. We 
>>> upgraded then restarted the HV1. 
>>> Then I edited the second VM2 in order to remove card from HV1. Then I tried 
>>> to add the card from HV2 to this VM2: the list of "host devices" is empty. 
>>> I started the VM1: it works fine (I didn't change anything on it). 
>> 
>> Hello! 
>> 
>> I'm afraid you have hit the bug [1]. For hosts where you need host 
>> devices, the best solution is most likely not upgrading them to 3.6.6 
>> for now. 
>> 
>> [1] https://bugzilla.redhat.com/show_bug.cgi?id=1341299 
>> 
>>> 
>>> -- 
>>> 
>>> 
>>> 
>>> 
>>> 
>>> Jérôme Chapelle 
>>> C entre I nformatique N ational de l' E nseignement S upérieur 
>>> 950 rue de Saint Priest 
>>> 34 097 M ONTPELLIER Cedex 5 
>>> 
>>> Tél. 04 67 14 14 14 
>>> eMail chape...@cines.fr 
>>> 
>> 
>>> ___ 
>

[ovirt-users] PCI Passthrough problem

2016-06-07 Thread Jérôme Chapelle
We are no more able to give a pci device to one VM. 
Here is the process we used to assign one device to one VM: 
- Click on the VM, and then on the tab "Host devices" 
- In the tab, click on "pin to host" and select the host that will give the pci 
device 
- Click on "add device", in the window appear all the pci devices: choose the 
pci card you wish (ex. pci__83_00_0 pci__83_00_1) and then click on the 
gray arrow to assign this device to the VM 
- Click on ok 
- boot the VM. 
The problem is that in the window that appears, there is not a single pci 
device listed (there should be many listed). 
I checked first: the IOMMU setting is set in the kernel. 

Our environment: 
We have two HV running CentOs7. 
Each of them have three fiber channel cards. First card is connected to a disk 
array, second one to the first tape library and third one to another tape 
library. 
We used to give the second card to a VM1, and the third card to another VM2. 
Both VMs ran on one HV or on the other in the past: everything was fine. 
What happened: 
Today we add to upgrade one of the HV (HV1). Both VMs ran on this HV1. We 
upgraded then restarted the HV1. 
Then I edited the second VM2 in order to remove card from HV1. Then I tried to 
add the card from HV2 to this VM2: the list of "host devices" is empty. 
I started the VM1: it works fine (I didn't change anything on it). 


-- 


    


Jérôme Chapelle 
C entre I nformatique N ational de l' E nseignement S upérieur 
950 rue de Saint Priest 
34 097 M ONTPELLIER Cedex 5 

Tél.04 67 14 14 14 
eMail   chape...@cines.fr 

___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users