6:35 a.m.
On 09/23/2013 02:23 PM, Rickard Kristiansson wrote:
i think this could be done with puppet, afaik some work for integrating it to
oVirt been done lately, though not sure if it's already available,
oved?
--
Michael Pasternak
RedHat, ENG-Virtualization R&D
6:41 a.m.
New subject: [Users] Multiple network interfaces on engine?
------=_Part_186583_54263086.1379936476704
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
OK. Maybe I'm asking for something irrelevant, but to answer Ren=C3=A9's qu=
estion; The use case is that there are two separate physical networks for g=
eneral IP and storage IP, and Engine seems to require to be able to reach t=
he NFS server to validate the connection when configuring mapping. So to be=
able to configure a NFS server, the Engine needs to be connected also to t=
he storage network.=20
It worked fine after manually tweaking the sysconfig/network-scripts, but t=
his config was then shortly overridden by Engine discovering and trying to =
use the second NIC for bonding with ovirtmgmt.=20
----- Ursprungligt meddelande -----
Fr=C3=A5n: "Michael Pasternak" <mpastern(a)redhat.com>=20
Till: "Rickard Kristiansson" <rickard(a)sanuksystems.com>, users(a)ovirt.org,
"=
Oved Ourfalli" <ovedo(a)redhat.com>=20
Kopia: "Ohad Levy" <olevy(a)redhat.com>=20
Skickat: m=C3=A5ndag, 23 sep 2013 13:35:40=20
=C3=84mne: Re: [Users] Multiple network interfaces on engine?=20
On 09/23/2013 02:23 PM, Rickard Kristiansson wrote:=20
Hmm.. yes, I do have the storage network configured in DC and
Cluster, an=
d also attached to all my hypervisors.=20
The question is just how I also attach it to the Engine server, as
this i=
sn't visible as a node in the GUI?=20
i think this could be done with puppet, afaik some work for integrating it =
to=20
oVirt been done lately, though not sure if it's already available,=20
oved?=20
---------------------------------------------------------------------------=
------------------------=20
at>=20
gure the networks on the engine?=20
> For nodes and AOI it's OK, but in my case I am running the
engine as a s=
eparate server which is not a node and does not show up as a Host in
the GU=
I.=20
> Or do I need to add the engine as a host in the GUI as well, to
be able =
to configure it?=20
C/s (if
you want it to be=20
atomic operation, i.e you expect connectivity loss during the
operation/s=
, setupnetworks should be used,=20
---------------------------------------------------------------------------=
-------------------------=20
Engine (3.3), where one is for ovirtmgmt and one is for a separate storage=
network.=20
>> It works fine for a minute or so after rebooting Engine, but
then the c=
onfiguration is changed by Engine, connecting both interfaces to bond0
and =
losing connectivity (as=20
>> obviously you can't bond those two interfaces when they
are connected t=
o separate physical networks..).=20
ine so that we can use separate networks on the network interfaces..?=20
>=20
> attaching logs would help, but i guess it happen cause you did not saved=
=20
> your host network config, i.e when host got rebooted, it performs
roll-b=
ack=20
--=20
Michael Pasternak=20
RedHat, ENG-Virtualization R&D=20
------=_Part_186583_54263086.1379936476704
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable
<html><body><div style=3D"font-family: Arial, Helvetica, sans-serif;
font-s=
ize: 12pt; color: #000000"><div>OK. Maybe I'm asking for something
irreleva=
nt, but to answer Ren=C3=A9's question; The use case is that there are two =
separate physical networks for general IP and storage IP, and Engine seems =
to require to be able to reach the NFS server to validate the connection wh=
en configuring mapping. So to be able to configure a NFS server, the Engine=
needs to be connected also to the storage network.<br></div><div>It
worked=
fine after manually tweaking the sysconfig/network-scripts, but this confi=
g was then shortly overridden by Engine discovering and trying to use the s=
econd NIC for bonding with ovirtmgmt.
<br></div><div><br></div><hr id=3D"zw=
chr"><div
style=3D"color:#000;font-weight:normal;font-style:normal;text-dec=
oration:none;font-family:Helvetica,Arial,sans-serif;font-size:12pt;" data-m=
ce-style=3D"color: #000; font-weight: normal; font-style: normal; text-deco=
ration: none; font-family: Helvetica,Arial,sans-serif; font-size: 12pt;"><b=
Fr=C3=A5n: </b>"Michael Pasternak"
<mpastern@redhat.com><br><b>Till:=
</b>"Rickard Kristiansson" &lt;rickard(a)sanuksystems.com&gt;,
users(a)ovirt.o=
rg, "Oved Ourfalli" &lt;ovedo(a)redhat.com&gt;<br><b>Kopia:
</b>"Ohad Levy" &=
lt;olevy(a)redhat.com&gt;<br><b>Skickat: </b>m=C3=A5ndag, 23 sep 2013
13:35:4=
0<br><b>=C3=84mne: </b>Re: [Users] Multiple network interfaces on
engine?<b=
r><div><br></div>On 09/23/2013 02:23 PM, Rickard Kristiansson
wrote:<br>>=
; Hmm.. yes, I do have the storage network configured in DC and Cluster, an=
d also attached to all my hypervisors.<br>> The question is just how I a=
lso attach it to the Engine server, as this isn't visible as a node in the =
GUI?<br><div><br></div>i think this could be done with puppet,
afaik some w=
ork for integrating it to<br>oVirt been done lately, though not sure if it'=
s already
available,<br><div><br></div>oved?<br><div><br></div>>
<br>>=
; <br>> ----------------------------------------------------------------=
---------------------------------------------------------------------------=
---------------------------------<br>> *Fr=C3=A5n: *"Michael
Pasternak" =
&lt;mpastern(a)redhat.com&gt;<br>&gt; *Till: *"Rickard
Kristiansson" <rick=
ard(a)sanuksystems.com&gt;<br>&gt; *Kopia: *users(a)ovirt.org,
"\"Ren=C3=A9 \"K=
och (ovido)\"\"" &lt;r.koch(a)ovido.at&gt;<br>&gt;
*Skickat: *m=C3=A5ndag, 23=
sep 2013 13:20:06<br>> *=C3=84mne: *Re: [Users] Multiple network interf=
aces on engine?<br>> <br>> On 09/23/2013 01:50 PM, Rickard
Kristianss=
on wrote:<br>>> Thanks
Michael,<br>>><br>>> I'm still try=
ing to get my head around this, but I can't see how I configure the network=
s on the engine?<br>>> For nodes and AOI it's OK, but in my case I
am=
running the engine as a separate server which is not a node and does not s=
how up as a Host in the GUI.<br>>> Or do I need to add the engine as
=
a host in the GUI as well, to be able to configure it?<br>>
<br>> IIU=
C your question, you have various options:<br>> <br>> 1.
GUI<br>> =
2. SDKs<br>> 3. CLI<br>> <br>> you need to create
network in DC, a=
ttach it to cluster, attach to host NIC/s (if you want it to be<br>> ato=
mic operation, i.e you expect connectivity loss during the operation/s, set=
upnetworks should be used,<br>> otherwise you can use attach-network dia=
logues in UI)<br>>
<br>>><br>>><br>>>
----------------=
---------------------------------------------------------------------------=
---------------------------------------------------------------------------=
------<br>>> *Fr=C3=A5n: *"Michael Pasternak"
&lt;mpastern(a)redhat.com=
><br>>> *Till: *"Rickard Kristiansson"
&lt;rickard(a)sanuksystems.co=
m><br>>> *Kopia: *users(a)ovirt.org<br>&gt;&gt;
*Skickat: *m=C3=A5nd=
ag, 23 sep 2013 12:21:40<br>>> *=C3=84mne: *Re: [Users] Multiple
netw=
ork interfaces on engine?<br>>><br>>> On
09/23/2013 12:54 PM, R=
ickard Kristiansson wrote:<br>>>>
Hi,<br>>>><br>>>&=
gt; I want to use two network interfaces connected to different networks in=
Engine (3.3), where one is for ovirtmgmt and one is for a separate storage=
network.<br>>>> It works fine for a minute or so after
rebooting =
Engine, but then the configuration is changed by Engine, connecting both in=
terfaces to bond0 and losing connectivity (as<br>>>> obviously
you=
can't bond those two interfaces when they are connected to separate physic=
al networks..).<br>>>> What is the mechanism doing
this,<br>>&g=
t;<br>>> please see this
[1].<br>>><br>>> [1] http://www.=
ovirt.org/Features/Design/Network/SetupNetworks<br>>>...
and it is possible to disable this automatic network bonding on the engine=
so that we can use separate networks on the network
interfaces..?<br>>&=
gt;<br>>> attaching logs would help, but i guess it happen cause you
=
did not saved<br>>> your host network config, i.e when host got
reboo=
ted, it performs roll-back<br>>> to old config, this is actually
disa=
ster recovery
mechanism.<br>>><br>>>><br>>>>
Rickar=
d<br>>>><br>>>>
**<br>>>><br>>>><br>>=
;>>
_______________________________________________<br>>>> U=
sers mailing list<br>>>>
Users(a)ovirt.org<br>&gt;&gt;&gt; http://li=
sts.ovirt.org/mailman/listinfo/users<br>>>><...
;<br>>> --<br>>><br>>> Michael
Pasternak<br>>> RedH=
at, ENG-Virtualization R&D<br>>><br>>
<br>> <br>> -- <br=
> <br>> Michael Pasternak<br>>
RedHat, ENG-Virtualization R&D=
<br>>
<br><div><br></div><br>--
<br><div><br></div>Michael Pasternak<br>=
RedHat, ENG-Virtualization
R&D<br></div><div><br></div></div></body></h=
tml>
------=_Part_186583_54263086.1379936476704--
8:05 a.m.
New subject: [Users] Multiple network interfaces on engine?
On 09/23/2013 02:41 PM, Rickard Kristiansson wrote:
engine does not need to be connected to storage network. it does all
storage access via the hosts.
the engine isn't trying to bond by itself. have you bonded the
interfaces on the host setup-network dialog?
8:17 a.m.
New subject: [Users] Multiple network interfaces on engine?
This is a multi-part message in MIME format.
--------------010701040300070806060900
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit
Rickard Kristiansson wrote:
I have also more than one
interface in my Engine but never had this
problem, I don't have a ovirtmgmt network on my Engine hosts thats not
needed. I can understand the need for the engine server to have a nic/ip
in the managment network AND a nic/ip in the storage network. I also
have a separate access/public network and ilo network.
It looks like you have a bond defined somewhere and it looks like you
have vdsm running on your engine server.
Just looking at my two engine servers and none has more than one ip. The
one in the test setup doesn't even have storage assigned and doesn't
need it. It sends mount commands to the hosts which DO have a separate
storage network. My engine hosts don't host the iso/export domains so
don't need to be in the storage network.
If you have the possibility, go back to a bare machine with
NetworkManager turned off and network turned on. Make sure networking is
working as expected, eth0=ip in managment, eth1= ip in storage.
Install/setup ovirt-engine-setup, don't install vdsm (gets installed if
you do allinone!) and things should 'just' work.
Regards,
Joop
--------------010701040300070806060900
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=UTF-8"
http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Rickard Kristiansson wrote:
<blockquote
cite="mid:1466620074.186584.1379936476705.JavaMail.root@sanuksystems.com"
type="cite">
<div
style="font-family: Arial,Helvetica,sans-serif; font-size: 12pt; color: rgb(0, 0,
0);">
<div>OK. Maybe I'm asking for something irrelevant, but to answer
René's question; The use case is that there are two separate physical
networks for general IP and storage IP, and Engine seems to require to
be able to reach the NFS server to validate the connection when
configuring mapping. So to be able to configure a NFS server, the
Engine needs to be connected also to the storage network.<br>
</div>
<div>It worked fine after manually tweaking the
sysconfig/network-scripts, but this config was then shortly overridden
by Engine discovering and trying to use the second NIC for bonding with
ovirtmgmt. <br>
</div>
</div>
</blockquote>
I have also more than one interface in my Engine but never had this
problem, I don't have a ovirtmgmt network on my Engine hosts thats not
needed. I can understand the need for the engine server to have a
nic/ip in the managment network AND a nic/ip in the storage network. I
also have a separate access/public network and ilo network.<br>
It looks like you have a bond defined somewhere and it looks like you
have vdsm running on your engine server.<br>
<br>
Just looking at my two engine servers and none has more than one ip.
The one in the test setup doesn't even have storage assigned and
doesn't need it. It sends mount commands to the hosts which DO have a
separate storage network. My engine hosts don't host the iso/export
domains so don't need to be in the storage network.<br>
<br>
If you have the possibility, go back to a bare machine with
NetworkManager turned off and network turned on. Make sure networking
is working as expected, eth0=ip in managment, eth1= ip in storage.
Install/setup ovirt-engine-setup, don't install vdsm (gets installed if
you do allinone!) and things should 'just' work.<br>
<br>
Regards,<br>
<br>
Joop<br>
<br>
</body>
</html>
--------------010701040300070806060900--
8:55 a.m.
New subject: [Users] Multiple network interfaces on engine?
------=_Part_188135_532733164.1379944506657
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
OK, thanks. Then I'm happy with the explanations that it is supposed to work the way I
initially expected..
That all operations are actually carried out on the nodes is good news, and should solve
all worries.
I'll go back and see what is actually going wrong on the engine and try to fix it.
I'd love to reinstall the engine, but as I can't find any documentation on how to
reliably move configs from one engine to another, and I was brave enough to already put
ovirt into production with a number of hypervisors, I'm somewhat limited to what I
can/dare to do.
----- Ursprungligt meddelande -----
I have also more than one interface in my Engine but never had this problem, I don't
have a ovirtmgmt network on my Engine hosts thats not needed. I can understand the need
for the engine server to have a nic/ip in the managment network AND a nic/ip in the
storage network. I also have a separate access/public network and ilo network.
It looks like you have a bond defined somewhere and it looks like you have vdsm running on
your engine server.
Just looking at my two engine servers and none has more than one ip. The one in the test
setup doesn't even have storage assigned and doesn't need it. It sends mount
commands to the hosts which DO have a separate storage network. My engine hosts don't
host the iso/export domains so don't need to be in the storage network.
If you have the possibility, go back to a bare machine with NetworkManager turned off and
network turned on. Make sure networking is working as expected, eth0=ip in managment,
eth1= ip in storage. Install/setup ovirt-engine-setup, don't install vdsm (gets
installed if you do allinone!) and things should 'just' work.
Regards,
Joop
_______________________________________________
Users mailing list
Users(a)ovirt.org
http://lists.ovirt.org/mailman/listinfo/users
------=_Part_188135_532733164.1379944506657
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: quoted-printable
<html><body><div style=3D"font-family: Arial, Helvetica, sans-serif;
font-s=
ize: 12pt; color: #000000"><div>OK, thanks. Then I'm happy with the
explana=
tions that it is supposed to work the way I initially
expected..<br></div><=
div>That all operations are actually carried out on the nodes is good news,=
and should solve all
worries.<br></div><div><br></div><div>I'll go back
an=
d see what is actually going wrong on the engine and try to fix it.<br></di=
v><div>I'd love to reinstall the engine, but as I can't find any
documentat=
ion on how to reliably move configs from one engine to another, and I was b=
rave enough to already put ovirt into production with a number of hyperviso=
rs, I'm somewhat limited to what I can/dare to
do.<br></div><div><br></div>=
<div><br></div><hr id=3D"zwchr"><div
data-mce-style=3D"color: #000; font-we=
ight: normal; font-style: normal; text-decoration: none; font-family: Helve=
tica,Arial,sans-serif; font-size: 12pt;"><br></div><div
data-mce-style=3D"c=
olor: #000; font-weight: normal; font-style: normal; text-decoration: none;=
font-family: Helvetica,Arial,sans-serif; font-size:
12pt;"><br></div><div =
data-mce-style=3D"color: #000; font-weight: normal; font-style: normal; tex=
t-decoration: none; font-family: Helvetica,Arial,sans-serif; font-size: 12p=
t;">I have also more than one interface in my Engine but never had this pro=
blem, I don't have a ovirtmgmt network on my Engine hosts thats not needed.=
I can understand the need for the engine server to have a nic/ip in the ma=
nagment network AND a nic/ip in the storage network. I also have a separate=
access/public network and ilo network.<br> It looks like you have a bond d=
efined somewhere and it looks like you have vdsm running on your engine ser=
ver.<br> <br> Just looking at my two engine servers and none has more than =
one ip. The one in the test setup doesn't even have storage assigned and do=
esn't need it. It sends mount commands to the hosts which DO have a separat=
e storage network. My engine hosts don't host the iso/export domains so don=
't need to be in the storage network.<br> <br> If you have the
possibility,=
go back to a bare machine with NetworkManager turned off and network turne=
d on. Make sure networking is working as expected, eth0=3Dip in managment, =
eth1=3D ip in storage. Install/setup ovirt-engine-setup, don't install vdsm=
(gets installed if you do allinone!) and things should 'just' work.<br>
<b=
r> Regards,<br> <br> Joop<br>
<br><div><br></div>__________________________=
_____________________<br>Users mailing
list<br>Users@ovirt.org<br>http://li=
sts.ovirt.org/mailman/listinfo/users<br></div><div><...
/html>
------=_Part_188135_532733164.1379944506657--
4222
days inactive
4222
days old
6 comments
5 participants
participants (5)
-
Itamar Heim -
Joop -
Michael Pasternak -
Oved Ourfalli -
Rickard Kristiansson