[ovirt-users] Re: VM support for Infiniband

2019-05-14 Thread Kevin Hung

On 9/30/2016 10:10 AM, Kevin Hung wrote:
oVirt sees the HCA as SR-IOV capable and it sees the VFs, so I believe 
I have set up SR-IOV correctly. I was able to do a generic passthrough 
of the HCA, but I was not able to pass through the VFs to more than 
one VM probably because they are all in the same IOMMU group. I first 
tried following the oVirt 3.6 Deep Dive video on SR-IOV [1] but was 
stuck because of the bridging problem.

[1] https://www.youtube.com/watch?v=A-MROZ8D06Y


I was able to pass through the VFs after fixing the IOMMU issue. I had 
to read up on IOMMU groups and discovered that the PCIe root ports on 
Xeon E3 processors do not support ACS [1], which is needed for the VFs 
to be in different IOMMU groups. I was able to plug the HCA into a 
different PCIe slot which was connected to the PCH instead, allowing 
VFIO to see the VFs in different IOMMU groups.


The way SR-IOV works for Ethernet network adapters in oVirt seems to be 
really nice as the scheduler takes care of the pass-through. Is this 
possible for Infiniband HCAs or is that a technical/architectural 
limitation?


[1] http://vfio.blogspot.com.ar/2014/08/iommu-groups-inside-and-out.html

Thanks,
Kevin Hung
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users

--
IMPORTANT!
This message has been scanned for viruses and phishing links.
However, it is your responsibility to evaluate the links and attachments you 
choose to click.
If you are uncertain, we always try to help.
Greetings helpd...@actnet.se



--
IMPORTANT!
This message has been scanned for viruses and phishing links.
However, it is your responsibility to evaluate the links and attachments you 
choose to click.
If you are uncertain, we always try to help.
Greetings helpd...@actnet.se

___
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-le...@ovirt.org
Privacy Statement: https://www.ovirt.org/site/privacy-policy/
oVirt Code of Conduct: 
https://www.ovirt.org/community/about/community-guidelines/
List Archives: 
https://lists.ovirt.org/archives/list/users@ovirt.org/message/2FYARIOQW4PQXA7TWJHY5AR4KEJ52ANI/


Re: [ovirt-users] VM support for Infiniband

2016-10-01 Thread Kevin Hung

On 9/30/2016 10:10 AM, Kevin Hung wrote:
oVirt sees the HCA as SR-IOV capable and it sees the VFs, so I believe 
I have set up SR-IOV correctly. I was able to do a generic passthrough 
of the HCA, but I was not able to pass through the VFs to more than 
one VM probably because they are all in the same IOMMU group. I first 
tried following the oVirt 3.6 Deep Dive video on SR-IOV [1] but was 
stuck because of the bridging problem.

[1] https://www.youtube.com/watch?v=A-MROZ8D06Y


I was able to pass through the VFs after fixing the IOMMU issue. I had 
to read up on IOMMU groups and discovered that the PCIe root ports on 
Xeon E3 processors do not support ACS [1], which is needed for the VFs 
to be in different IOMMU groups. I was able to plug the HCA into a 
different PCIe slot which was connected to the PCH instead, allowing 
VFIO to see the VFs in different IOMMU groups.


The way SR-IOV works for Ethernet network adapters in oVirt seems to be 
really nice as the scheduler takes care of the pass-through. Is this 
possible for Infiniband HCAs or is that a technical/architectural 
limitation?


[1] http://vfio.blogspot.com.ar/2014/08/iommu-groups-inside-and-out.html

Thanks,
Kevin Hung
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


Re: [ovirt-users] VM support for Infiniband

2016-09-30 Thread Kevin Hung

On 9/30/2016 9:28 AM, Michal Skrivanek wrote:

On 30 Sep 2016, at 03:56, Kevin Hung <kh...@nullaxiom.com> wrote:

Hello,

I have been trying (unsuccessfully) to add my Infiniband HCAs on my VM hosts to 
oVirt 4.0.4. I am not using IBoIP, just plain Infiniband. I would like to be 
able to pass through the card (or better yet the VF from SR-IOV) to the guest 
VM.

SR-IOV should work (or general PCI passthrough), it’s then transparent to 
oVirt. Did you not succeed with setting up SR-IOV or generic passthrough of the 
HCA?
oVirt sees the HCA as SR-IOV capable and it sees the VFs, so I believe I 
have set up SR-IOV correctly. I was able to do a generic passthrough of 
the HCA, but I was not able to pass through the VFs to more than one VM 
probably because they are all in the same IOMMU group. I first tried 
following the oVirt 3.6 Deep Dive video on SR-IOV [1] but was stuck 
because of the bridging problem.





 From what I can figure out, oVirt is trying to bridge the interface but I 
don't think that will work as shown here [1]. Does oVirt currently have support 
for passing through Infiniband to VMs? If not, are there any plans to support 
it?

Well, at least for generic passthrough the PCI device would be detached from 
the host so should not be recognized/touched by oVirt networking code at all.

Thanks,
michal


[1] http://lists.ovirt.org/pipermail/users/2016-July/040898.html

Thanks,
Kevin Hung
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users



[1] https://www.youtube.com/watch?v=A-MROZ8D06Y
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-07-04 Thread Kevin Hung


On 7/4/2016 4:38 AM, Simone Tiraboschi wrote:

On Sun, Jul 3, 2016 at 5:57 AM, Kevin Hung <kh...@nullaxiom.com> wrote:

4. Using cloud-init with default values to customize the engine appliance
can fail on the line "Creating/refreshing DWH database schema" if it takes
longer than 600 seconds to return output. This may apply to any other step
that takes a long time to complete. The VM no longer appears to be exist
after the setup exits that so I am unable to debug.

600 seconds seams more than a reasonable time to create an empty DB,
if it requires more than 10 minutes for a simple/short operation there
is probably something strange with the storage.
I monitored the host RAM/CPU usage and the utilization of the Ethernet 
interface on the shared storage. RAM and CPU usage were minimal and 
there was barely anything going through the network interface. I can 
confirm that the network interface is fine as it is heavily utilized 
when the engine setup uses it to copy the image to the storage. I'm not 
sure what other statistics I should monitor to see if there's a bottleneck.

5. Without using cloud-init, the setup creates an engine VM that I cannot
log into (it does not seem to use the engine admin password or a blank
password).

Yes, the engine VM host-name and its root password are configured via
cloud-init and there is not default password.
If you want to avoid using cloud-init you have to reset the root
password of the engine VM as for any el7 machine.


6. Destroying the VM (option 4) leaves the files intact on the shared
storage so I cannot restart setup without deleting those first. This may be
intentional, but the use of kvm terminology (destroy for power off) is not
common, not to mention that "virsh -r list --all" does not list the VM
anymore.

On failures, there is not just the engine VM disk but a whole storage
domain for hosted-engine which also contains ancillary disks.
Re-deploying over a dirty storage is not supported so please clean up
the whole storage domain on failures.


7. Unable to deploy second host through web UI (error "Failed to configure
management network on host node2 due to setup networks failure.") or using

This is not hosted-engine specific:
https://bugzilla.redhat.com/show_bug.cgi?id=1350763
Thanks for pointing out the BZ ticket. I'm not entirely certain that's 
the same issue I was seeing. Of course, I have no way of verifying 
anymore as I have already re-deployed using CentOS instead of Node.



ovirt-hosted-engine-setup (it looks like it can't connect to or doesn't
start the broker service).
8. Random errors to stderr: "vcpu0 unhandled rdmsr" (this seems to be an

Are you running in a nested env?

No, this is bare-metal, not a nested environment.
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users


Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-07-02 Thread Kevin Hung
Looks like there still needs to be some work done on oVirt 4.0 Node and 
ovirt-hosted-engine-setup before it's ready for general consumption. I 
have spent days trying to get this to work, and only got it running (on 
one host) after encountering 8 serious issues (7 below and the initial 
glusterfs one). I have not been able to successfully deploy a second 
host (see issue 7 below). I will be moving back to deploying hosts using 
CentOS (with either oVirt 4.0 or oVirt 3.6) as I need a working oVirt 
deployment up and running.


In case anyone is interested in reproducing the issues, I used the Node 
ISO here [1] and the latest (7/2/2016) engine appliance OVA here [2]. 
Those seem to be the "official" files as far as I can tell (which is 
difficult as the documentation is not clear).


List of issues:
1. The error I mentioned seems to be an problem with the code. I 
bypassed it by deleting 
/usr/libexec/vdsm/hooks/before_network_setup/50_fcoe.
2. ovirt-hosted-engine-setup is unable to connect to the vdsm service if 
the FQDN of the node is not resolvable (i.e. if a DNS server is not 
entered in the initial setup). This should be checked in either the 
initial oVirt Node setup process or the beginning of 
ovirt-hosted-engine-setup.
3. The management bridge does not get created properly when the server 
is set up with a manually configured DNS server and running 
NetworkManager (the default on Node). It seems like a bug has been filed 
for this back in 2014. [3]
4. Using cloud-init with default values to customize the engine 
appliance can fail on the line "Creating/refreshing DWH database schema" 
if it takes longer than 600 seconds to return output. This may apply to 
any other step that takes a long time to complete. The VM no longer 
appears to be exist after the setup exits that so I am unable to debug.
5. Without using cloud-init, the setup creates an engine VM that I 
cannot log into (it does not seem to use the engine admin password or a 
blank password).
6. Destroying the VM (option 4) leaves the files intact on the shared 
storage so I cannot restart setup without deleting those first. This may 
be intentional, but the use of kvm terminology (destroy for power off) 
is not common, not to mention that "virsh -r list --all" does not list 
the VM anymore.
7. Unable to deploy second host through web UI (error "Failed to 
configure management network on host node2 due to setup networks 
failure.") or using ovirt-hosted-engine-setup (it looks like it can't 
connect to or doesn't start the broker service).
8. Random errors to stderr: "vcpu0 unhandled rdmsr" (this seems to be an 
upstream bug) and "multipath: error getting device" (this has been an 
issue for years with oVirt and seems to be due to multipathing being on 
by default even for systems where that does not apply).


[1] 
http://resources.ovirt.org/pub/ovirt-4.0/iso/ovirt-node-ng-installer/ovirt-node-ng-installer-ovirt-4.0-2016062412.iso
[2] 
http://jenkins.ovirt.org/view/All/job/ovirt-appliance_ovirt-4.0_build-artifacts-el7-x86_64/

[3] https://bugzilla.redhat.com/show_bug.cgi?id=1160423

On 7/1/2016 8:37 PM, Kevin Hung wrote:
It looks like I'm now getting an error when the deployment tries to 
configure the management bridge.


Setup log:

2016-07-01 20:29:47 INFO otopi.plugins.gr_he_common.network.bridge 
bridge._misc:

372 Configuring the management bridge
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc
:384 networks: {'ovirtmgmt': {'nic': 'eno1', 'ipaddr': 
u'192.168.1.211', 'netmask': u'255.255.255.0', 'bootproto': u'none', 
'gateway': u'192.168.1.1', 'defaultRoute': True}}
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc

:385 bonds: {}
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc

:386 options: {'connectivityCheck': False}
2016-07-01 20:29:48 DEBUG otopi.context context._executeMethod:142 
method exception

Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/otopi/context.py", line 132, 
in _executeMethod

method['method']()
  File 
"/usr/share/ovirt-hosted-engine-setup/scripts/../plugins/gr-he-common/network/bridge.py", 
line 387, in _misc

_setupNetworks(conn, networks, bonds, options)
  File 
"/usr/share/ovirt-hosted-engine-setup/scripts/../plugins/gr-he-common/network/bridge.py", 
line 405, in _setupNetworks

'message: "%s"' % (networks, code, message))
RuntimeError: Failed to setup networks {'ovirtmgmt': {'nic': 'eno1', 
'ipaddr': u'192.168.1.211', 'netmask': u'255.255.255.0', 'bootproto': 
u'none', 'gateway': u'192.168.1.1', 'defaultRoute': True}}. Error 
code: "78" message: "Hook error: Hook Error: ('Traceback (most recent 
call last):\n  File 
"/usr/libexec/vdsm/hooks/before_network_setup/50_fcoe", line 18, in 
\nfrom vdsm.netconfpersistence import 
RunningCon

Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-07-01 Thread Kevin Hung
It looks like I'm now getting an error when the deployment tries to 
configure the management bridge.


Setup log:

2016-07-01 20:29:47 INFO otopi.plugins.gr_he_common.network.bridge 
bridge._misc:

372 Configuring the management bridge
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc
:384 networks: {'ovirtmgmt': {'nic': 'eno1', 'ipaddr': u'192.168.1.211', 
'netmask': u'255.255.255.0', 'bootproto': u'none', 'gateway': 
u'192.168.1.1', 'defaultRoute': True}}
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc

:385 bonds: {}
2016-07-01 20:29:48 DEBUG otopi.plugins.gr_he_common.network.bridge 
bridge._misc

:386 options: {'connectivityCheck': False}
2016-07-01 20:29:48 DEBUG otopi.context context._executeMethod:142 
method exception

Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/otopi/context.py", line 132, 
in _executeMethod

method['method']()
  File 
"/usr/share/ovirt-hosted-engine-setup/scripts/../plugins/gr-he-common/network/bridge.py", 
line 387, in _misc

_setupNetworks(conn, networks, bonds, options)
  File 
"/usr/share/ovirt-hosted-engine-setup/scripts/../plugins/gr-he-common/network/bridge.py", 
line 405, in _setupNetworks

'message: "%s"' % (networks, code, message))
RuntimeError: Failed to setup networks {'ovirtmgmt': {'nic': 'eno1', 
'ipaddr': u'192.168.1.211', 'netmask': u'255.255.255.0', 'bootproto': 
u'none', 'gateway': u'192.168.1.1', 'defaultRoute': True}}. Error code: 
"78" message: "Hook error: Hook Error: ('Traceback (most recent call 
last):\n  File "/usr/libexec/vdsm/hooks/before_network_setup/50_fcoe", 
line 18, in \nfrom vdsm.netconfpersistence import 
RunningConfig\nImportError: No module named netconfpersistence\n',)"
2016-07-01 20:29:48 ERROR otopi.context context._executeMethod:151 
Failed to execute stage 'Misc configuration': Failed to setup networks 
{'ovirtmgmt': {'nic': 'eno1', 'ipaddr': u'192.168.1.211', 'netmask': 
u'255.255.255.0', 'bootproto': u'none', 'gateway': u'192.168.1.1', 
'defaultRoute': True}}. Error code: "78" message: "Hook error: Hook 
Error: ('Traceback (most recent call last):\n File 
"/usr/libexec/vdsm/hooks/before_network_setup/50_fcoe", line 18, in 
\nfrom vdsm.netconfpersistence import 
RunningConfig\nImportError: No module named netconfpersistence\n',)"



On 7/1/2016 5:21 PM, Kevin Hung wrote:
Thank you Sahina, that was the issue. I upgraded my glusterfs server 
to 3.7.11 and I was able to continue with the deployment. I am seeing 
other issues with deployment, but I will look into those myself first. 
Bug has been logged [1].


[1] https://bugzilla.redhat.com/show_bug.cgi?id=1352165


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


Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-07-01 Thread Kevin Hung
Thank you Sahina, that was the issue. I upgraded my glusterfs server to 
3.7.11 and I was able to continue with the deployment. I am seeing other 
issues with deployment, but I will look into those myself first. Bug has 
been logged [1].


[1] https://bugzilla.redhat.com/show_bug.cgi?id=1352165

On 7/1/2016 1:16 AM, Sahina Bose wrote:
This is a compatibility issue on executing cli commands. Is it 
possible to upgrade your glusterfs server to higher version, or 
downgrade the glusterfs-cli on node to 3.7.6.


Please log a bug in HE deploy to handle this error so that 
installation can proceed. (--remote-host option throws errors when 
there's a version mismatch)


On 07/01/2016 10:25 AM, Kevin Hung wrote:
Version 3.7.11 on node and version 3.7.6 on server. The node was set 
up using the ovirt-node-ng-installer-ovirt-4.0-2016062412 ISO.


On 7/1/2016 12:50 AM, Sahina Bose wrote:

Sorry, missed that you already ran it.
Version of glusterfs on node and server?




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


Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-06-30 Thread Kevin Hung
Version 3.7.11 on node and version 3.7.6 on server. The node was set up 
using the ovirt-node-ng-installer-ovirt-4.0-2016062412 ISO.


On 7/1/2016 12:50 AM, Sahina Bose wrote:

Sorry, missed that you already ran it.
Version of glusterfs on node and server?

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


Re: [ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-06-30 Thread Kevin Hung

The same as the other command, two blank lines.

On 7/1/2016 12:41 AM, Sahina Bose wrote:

What's the output of below from the node?
# gluster volume info mgmttank --remote-host=storage1.nullaxiom.com

On 07/01/2016 08:52 AM, Kevin Hung wrote:

Hello,

I thought I would ask the list if anyone is aware of this issue (or 
if I am doing something obviously wrong) before I submit a bug report.


It looks like I am not able to choose glusterfs as a storage option 
for deploying a Hosted Engine in oVirt 4.0 Node Next. I believe it is 
failing to execute the following command:
/sbin/gluster --mode=script --xml volume info mgmttank 
--remote-host=storage1.nullaxiom.com


When I run the command manually, I get a blank line as output and 
checking in /var/log/glusterfs/cli.log, it seems to be exiting with 
error code -2.


I have tried disabling the firewall on both host and gluster server, 
but it did not make a difference. I am able to manually mount the 
gluster volume on the host, and I had a working oVirt 3.6 
installation using the same exact gluster server.


Console output below:

[ INFO  ] Stage: Initializing
[ INFO  ] Generating a temporary VNC password.
[ INFO  ] Stage: Environment setup
  During customization use CTRL-D to abort.
  Continuing will configure this host for serving as 
hypervisor and create a VM where you have to install the engine 
afterwards.

  Are you sure you want to continue? (Yes, No)[Yes]:
  It has been detected that this program is executed through 
an SSH connection without using screen.
  Continuing with the installation may lead to broken 
installation if the network connection fails.
  It is highly recommended to abort the installation and run 
it inside a screen session using command "screen".

  Do you want to continue anyway? (Yes, No)[No]: yes
[ INFO  ] Hardware supports virtualization
  Configuration files: []
  Log file: 
/var/log/ovirt-hosted-engine-setup/ovirt-hosted-engine-setup-20160630225930-ef5wd4.log

  Version: otopi-1.5.0 (otopi-1.5.0-1.el7.centos)
[ INFO  ] Stage: Environment packages setup
[ INFO  ] Stage: Programs detection
[ INFO  ] Stage: Environment setup
[ INFO  ] Stage: Environment customization

  --== STORAGE CONFIGURATION ==--

  Please specify the storage you would like to use 
(glusterfs, iscsi, fc, nfs3, nfs4)[nfs3]: glusterfs
[ INFO  ] Please note that Replica 3 support is required for the 
shared storage.
  Please specify the full shared storage connection path to 
use (example: host:/path): storage1.nullaxiom.com:/mgmttank
[ ERROR ] Cannot access storage connection 
storage1.nullaxiom.com:/mgmttank: Command '/sbin/gluster' failed to 
execute
  Please specify the full shared storage connection path to 
use (example: host:/path):

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




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


[ovirt-users] Unable to set up oVirt 4.0 HE using glusterfs storage

2016-06-30 Thread Kevin Hung

Hello,

I thought I would ask the list if anyone is aware of this issue (or if I 
am doing something obviously wrong) before I submit a bug report.


It looks like I am not able to choose glusterfs as a storage option for 
deploying a Hosted Engine in oVirt 4.0 Node Next. I believe it is 
failing to execute the following command:
/sbin/gluster --mode=script --xml volume info mgmttank 
--remote-host=storage1.nullaxiom.com


When I run the command manually, I get a blank line as output and 
checking in /var/log/glusterfs/cli.log, it seems to be exiting with 
error code -2.


I have tried disabling the firewall on both host and gluster server, but 
it did not make a difference. I am able to manually mount the gluster 
volume on the host, and I had a working oVirt 3.6 installation using the 
same exact gluster server.


Console output below:

[ INFO  ] Stage: Initializing
[ INFO  ] Generating a temporary VNC password.
[ INFO  ] Stage: Environment setup
  During customization use CTRL-D to abort.
  Continuing will configure this host for serving as hypervisor 
and create a VM where you have to install the engine afterwards.

  Are you sure you want to continue? (Yes, No)[Yes]:
  It has been detected that this program is executed through an 
SSH connection without using screen.
  Continuing with the installation may lead to broken 
installation if the network connection fails.
  It is highly recommended to abort the installation and run it 
inside a screen session using command "screen".

  Do you want to continue anyway? (Yes, No)[No]: yes
[ INFO  ] Hardware supports virtualization
  Configuration files: []
  Log file: 
/var/log/ovirt-hosted-engine-setup/ovirt-hosted-engine-setup-20160630225930-ef5wd4.log

  Version: otopi-1.5.0 (otopi-1.5.0-1.el7.centos)
[ INFO  ] Stage: Environment packages setup
[ INFO  ] Stage: Programs detection
[ INFO  ] Stage: Environment setup
[ INFO  ] Stage: Environment customization

  --== STORAGE CONFIGURATION ==--

  Please specify the storage you would like to use (glusterfs, 
iscsi, fc, nfs3, nfs4)[nfs3]: glusterfs
[ INFO  ] Please note that Replica 3 support is required for the shared 
storage.
  Please specify the full shared storage connection path to use 
(example: host:/path): storage1.nullaxiom.com:/mgmttank
[ ERROR ] Cannot access storage connection 
storage1.nullaxiom.com:/mgmttank: Command '/sbin/gluster' failed to execute
  Please specify the full shared storage connection path to use 
(example: host:/path):

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


[ovirt-users] New oVirt Node releases?

2015-12-21 Thread Kevin Hung

Hello,

I have noticed that there have not been any official releases of oVirt 
Node since 3.5.2 [1]. I have tried looking through the Users and the 
Announce mailing lists for any information regarding the lack of 
releases, but I was not able to find anything from the past few months. 
The official documentation for oVirt seems to indicate that oVirt Node 
is still supported.


Are there just build problems preventing oVirt Node from being released? 
Or are we supposed to use either the latest Jenkins build or our own 
build [2] if we wish to deploy oVirt Node?


[1] http://resources.ovirt.org/pub/ovirt-3.5/iso/ovirt-node/el7-3.5.2/
[2] http://www.ovirt.org/Node_Building
___
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users