[ovirt-users] 3.6 issues when using an existing gluster volume for the hosted-engine

[Michael DePaulo]

Hi,

I am attempting to setup a Ovirt 3.6 (currently beta 1) on a test
environment with an existing (but empty) gluster volume for the hosted
engine.

I am currently experiencing 4 issues after I having deployed the
hosted-engine appliance VM on only 1 oVirt host:
1. The gluster volume is not added as a storage domain by the end of
the hosted-engine-setup.
2. When I try to import the gluster volume as a storage domain, I get
this error:
Error while executing action Attach Storage Domain: AcquireHostIdFailure
In the logs below, I had not attempted to import it yet though.
3. The hosted engine VM is not listed, although the host is listed and
its number of VMs is listed as "1". (I have seen recent emails/bug
reports about what seems to be the same issue.)
4. The hosted VM shut down automatically after several minutes.

I last ran ovirt-hosted-engine-setup on galactica on 2015-08-03 at
around 22:47 -0400

#4 was probably caused by the fact that during a previous attempt (on
2015-08-02) to setup the hosted-engine, I had
/var/lib/ovirt-hosted-engine-ha/ha.conf containing to
local_maintenance=True. That value was not changed when I ran
ovirt-hosted-engine-setup on 2015-08-03. Today, I ran the
hosted-engine command to disable maintenance, and the hosted-engine VM
started back up. I can access it over https, but the remaining issues
are still present.

----
More Details:

The 3 Gluster servers:
Server 1: galactica: RHEL 7.1 with Gluster 3.7.3
Server 2: death-star: RHEL 7.1 with Gluster 3.7.3
Server 3: mothership: Fedora 22 with Gluster 3.7.3

I only intend for servers 1 and 2 to be oVirt hosts at this time.

All servers have iptables disabled, are on the same LAN, have SELinux
set to enforcing, and have gluster configured with the hyper-converged
settings listed here:
http://www.ovirt.org/Features/Self_Hosted_Engine_Hyper_Converged_Gluster_Support

I have DNS forward and reverse lookups configured correctly for all 3
machines plus the hosted engine.

The existing gluster volume is "gv-virt", and it has the necessary
"options" for gluster to work with oVirt. Other than the .trashcan
folder, there were no files/folders on it when I ran
ovirt-hosted-engine-setup.

I am able to mount and write to gv-virt via mount.glusterfs.

Note that I have run hosted-engine-setup repeatedly on servers 1 and 2
as I try to resolve these issues. My cleanup between running it may
not have been correct, as exemplified by the local_maintenance
setting. I have been deleting the files on the gluster volume between
running it.

I have the relevant contents of the relevant directories under
/var/log/, /var/lib/, and /etc/ here:
https://drive.google.com/folderview?id=0B2L_OrCln9ShMWprWVlpZVVocGs&usp=sharing

I also put the `gluster volume status gv-virt` output there.

Also, these 2 volumes are currently mounted on galactica:
galactica.depaulo.org:/gv-virt                          200G  2.7G
198G   2% /rhev/data-center/mnt/glusterSD/galactica.depaulo.org:_gv-virt
/dev/loop1                                              2.0G  3.1M
1.9G   1% /rhev/data-center/mnt/_var_lib_ovirt-hosted-engine-setup_tmp2HOqSq

Thanks in advance,

-Mike