Red Hat Bugzilla – Bug 1326709
Unable to change ovirtmgmt network to another interface
Last modified: 2017-12-22 02:46:14 EST
Created attachment 1146823 [details] supervdsm.log Description of problem: During hosted-engine deployment on additional host, the following config was specified: OVEHOSTED_CORE/isAdditionalHost=bool:True OVEHOSTED_STORAGE/hostID=str:3 OVEHOSTED_FIRST_HOST/fetchAnswer=bool:False OVEHOSTED_NETWORK/bridgeIf=str:em1 OVEHOSTED_ENGINE/appHostName=str:rhsdev14 OVEHOSTED_ENGINE/adminPassword=str:redhat However, the ovirtmgmt interface was created on em2. This causes errors when HE runs on this host. Changing the ovirtmgmt to correct interface via UI does not work either. Logs attached Version-Release number of selected component (if applicable): How reproducible: Always on one server. Steps to Reproduce: As in description Additional info:
Created attachment 1146824 [details] engine.log
Meital - was this tested?
Sahina, can you please attach also hosted-engine-setup logs?
Oh, sorry, the issue is on additional host. On the first host, hosted-engine-setup directly creates the bridge before starting the engine VM; the bridge is created via setupNetwork on the interface specified with OVEHOSTED_NETWORK/bridgeIf On additional hosts, we simply call hosts.add so OVEHOSTED_NETWORK/bridgeIf is not relevant at all. AFAIK host-deploy will choose the interface where the engine can reach the host based on how the host address resolves.
Simone, yes. the host address that hosted-engine picked up - did resolve to em2. The additional host had 2 FQDNs - rhsdev14.lab..com and dhcp42-219.lab..com. Host deploy used dhcp42-219.lab..com which resolves to em2. Simone, is there a plan to allow specifying FQDN and interface for additional hosts? Also, I was not able to use Setup networks to change ovirtmgmt to em1 interface. That should have worked, IMO.
(In reply to Sahina Bose from comment #5) > Simone, is there a plan to allow specifying FQDN Yes, we have this one: https://bugzilla.redhat.com/show_bug.cgi?id=1251968 Now we are simply calling socket.getfqdn() > and interface for > additional hosts? Nope since we can pass an interface on hosts.add > Also, I was not able to use Setup networks to change ovirtmgmt to em1 > interface. That should have worked, IMO. Absolutely
Changing the title of this bug, as the initial issue is addressed by RFE
Sahina, I do not understand your request. ovitrmgmt is created on top of the interface that carries that IP address which the hostname resolves to. Do you want to change that? Why?
Dan, in my case the host had 2 fqdns : - rhsdev14.lab..com and dhcp42-219.lab..com Due to some setting, hostname for host resolved to dhcp42-219.lab..com (associated with em2) The HE deployment picked this but I want the em1 <-> ovirtmgmt. Using setup networks and changing ovirtmgmt to em1 did not work, as I mentioned. If the flow for this is 1. Edit hostname and change to rhsdev14 (associated with em1) from ovirt engine 2. Re-install host -- will this associate ovirtmgmt with em1 If it does, we can close this bug.
I'm afraid that there is no reasonable way to change the management address of a host after it has been added to Engine. The only solution I see is to remove the host, and add it explicitly with its requested name.
Moving this from 3.6 - this looks like one of the gotchas that users need to be aware of before deploying hosted engine - ensure that hostname on host resolves to required FQDN and em1.
It is very annoying, but fixing this runs deep, and I do not see it happening soon.