[Users] ovirt 3.2 networking setup problems

Dan Kenigsberg danken at redhat.com
Tue Feb 26 12:16:40 UTC 2013


On Tue, Feb 26, 2013 at 03:25:13PM +0400, Yuriy Demchenko wrote:
> i've made some additional research on problem #1:
> it seems that /usr/share/vdsm/configNetwork.py treats "gateway=null"
> value as true, so setting the error flag
> in "_addNetworkValidation" method there's a check at line 813:
> >        if netmask or gateway:
> >            raise ConfigNetworkError(ne.ERR_BAD_ADDR,
> >                        "Specified netmask or gateway but not ip")
> as "ipaddr", sent by engine = "", its check evaluated as false, then
> "gateway" evaluated, but it contains value (word "null"), so i
> presume evaluated as true and exception is raised.

I believe that this issue is due to code introduced by dreyou.
http://www.dreyou.org/ovirt/vdsm32.patch

diff --git a/vdsm/configNetwork.py b/vdsm/configNetwork.py
index 53debfa..c8a80f1 100755
--- a/vdsm/configNetwork.py
+++ b/vdsm/configNetwork.py
@@ -873,6 +873,11 @@ def addNetwork(network, vlan=None, bonding=None, nics=None, ipaddr=None,
     _netinfo = netinfo.NetInfo()
     bridged = utils.tobool(bridged)
 
+    # Hack here, netmask may be not defined, if this happen,
+    # set netmask to 255.255.255.0
+    if not netmask:
+        netmask = "255.255.255.0"
+
     if mtu:
         mtu = int(mtu)
 
I do not know the motivation for it - I hope that Dreyou himself can help.

Dreyou, others - I would very much appreciate it if you send such patches to
upstream gerrit.ovirt.org. They deserve discussion and hopefully -
incorporation into mainline vdsm.

> 
> I've added my notes to https://bugzilla.redhat.com/show_bug.cgi?id=911957
> 
> 
> Yuriy Demchenko
> 
> On 02/26/2013 12:37 PM, Yuriy Demchenko wrote:
> >Hi,
> >
> >I'm testing ovirt 3.2 on centos 6.3 (from dreyou repo) and have a
> >problems setting up networks:
> >1.
> >- clean install of ovirt-engine, added a host (6 NICs, only 1
> >configured with static ip before adding it to ovirt)
> >- set up bonding for ovirtmgmt from 2 NICs - em1+em2
> >- set up 2 additional networks ("trunk" with no vlanid, "vlan50"
> >with vlanid=50), attached it to cluster
> >- trying to set up host networks - assigned "trunk" to one of
> >unused NIC, "vlan50" to another, boot protocol in both - "none"
> >and when i'm trying to apply changes - i get "Specified netmask or
> >gateway but not ip"
> >What is wrong? why the hell it thinks i pass some netmask/gw in
> >first place? I dont need any ip configuration on this networks,
> >i'll be using them only for VM's
> >
> >2.
> >when i'm trying to team NIC's in bonds, i've got choice from only
> >3 bonds: "bond0", "bond1", "bond4" - why is that list is fixed?
> >If I've allready defined bond0+bond1 prior adding host to ovirt
> >(via kickstart), even if they aren't used - choice list in ovirt
> >host networks is short to "bond4" and trying to bond more
> >interfaces ends up with "there no bonds available"

Genady has reported this issue to me yesterday :-(.
It is fixed by http://gerrit.ovirt.org/12447 and deseves a rebuild of vdsm.
Please confirm that my patch solves this issue.




More information about the Users mailing list