
On Sun, Apr 6, 2014 at 4:55 PM, Yedidyah Bar David <didi@redhat.com> wrote:
----- Original Message -----
From: "Andrew Lau" <andrew@andrewklau.com> To: "users" <users@ovirt.org> Sent: Sunday, April 6, 2014 3:10:19 AM Subject: [Users] Hosted-Engine purpose for gateway check?
Hi,
I was recently playing around with the new ovirt 3.4 ga, I'm very happy all those issues I reported got fixed :D
I found a new issue regarding the use of PREFIX vs NETMASK which I've uploaded here https://bugzilla.redhat.com/show_bug.cgi?id=1084685
Anyway -- I'm wondering what is the purpose for the gateway check in the hosted-engine setup? In my test case, I had the following NIC configuration
eth0 - public (has gateway) eth1 - management eth1.1 - storage eth2 - vm data (no IP address)
So during the hosted engine install, it will not let me assign eth2 as the NIC because it has no IP address or gateway. So I proceed to use eth1 instead as, as it has an IP address but again that would fail because no gateway. Luckily I have a L3 switch, so I put up a gateway for eth1 and that solved that issue.
What is the gateway check supposed to achieve? I also tried to put in my eth0's IP address as the gateway but it still failed because of those config issues. If management/ovirtmgmt/vmdata are all on a L2 switch environment, effectively there becomes no gateway and it prevents the installation.
It actually does not need to really be a gateway. It's used only as part of a calculation trying to assess the liveliness of the host. See [1] for details, especially pages 33-34.
[1] http://www.ovirt.org/images/8/88/Hosted_Engine_Deep_Dive.pdf -- Didi
I've recall reading that pdf before - however my comments are a little aimed towards why does the setup require the GATEWAY=x be in the ifcfg-ethx file when it also asks for the gateway in the otopi setup. It seems a little redundant and also prevents the ability to proceed with the setup if you're in a L2 switch environment. The hosted-engine VM will require a gateway as it only has one nic and needs to be publicly accessible, so let's say we have: eth1 -> ovirtmgmt (172.16.0.10) -> hosted-engine (192.168.100.10 w/ 192.168.100.1 as gateway) Isn't 192.168.100.1 the gateway we want to be checking for? Although now that I think of it, I'm confused where the gateway check has it's example scenario, is it just for checking to make sure the hosted-engine will be externally accessible? Wouldn't it also work to do something like ethtool and check the link exists instead.