[ovirt-users] Recovering from an aborted hosted-engine --deploy

[Yedidyah Bar David]

----- Original Message -----
> From: "Bob Doolittle" <bdoolittle at teradici.com>
> To: "users" <users at ovirt.org>
> Sent: Friday, May 16, 2014 10:12:43 PM
> Subject: [ovirt-users] Recovering from an aborted hosted-engine --deploy
> 
> Hi,
> 
> I had an issue at the end of my hosted-engine --deploy.
> 
> My VM was stuck during OS installation because I was unable to configure
> the network for some reason.
> 
> So I chose the final option "3" to abort the deployment.
> 
> Now, I seem to be stuck. If I try to re-run --deploy it says it's
> already installed.
> 
> If I try to run --vm-status, it says:
> Cannot connect to the HA daemon, please check the logs.
> 
> But it's not clear what logs I should look at to get more info.
> 
> I've attached the most recently modified logs. I can't get useful
> information out of them I'm afraid.
> There is no log for /var/log/ovirt-hosted-engine-ha
> 
> Any clues?

There is currently no cleanup tool for hosted-engine.
You can remove all relevant packages (*ovirt*, *vdsm*, *libvirt*) and
conf files of them (/etc/*ovirt* /etc/*vdsm* /etc/*libvirt* /etc/pki/vdsm)
and then try again.

The sources of ovirt-hosted-engine-setup include a script cleanup.sh
that cleans up some things.

During development, and especially while working on bugzilla bug #1034634,
I did the following:
1. Keep in /root/clean-files copies of parts of /etc from a "clean"
installation (e.g. after deleting stuff as explained above and yum
install hosted engine setup but before deploy):
[root at didi-box1 clean-files]# tree
.
├── libvirt
│   ├── libvirtd.conf
│   ├── qemu.conf
│   └── qemu-sanlock.conf
├── logrotate.d
│   └── libvirtd
├── ovirt-hosted-engine-ha
│   └── broker.conf
└── sysconfig
    └── libvirtd

4 directories, 6 files

2. To clean up, I ran this script:
=========================================================================
#!/bin/sh -x

CLEAN=$HOME/clean-files

service vdsmd stop
service supervdsmd stop
initctl stop libvirtd

/bin/rm /etc/init/libvirtd.conf
/bin/cp $CLEAN/libvirt/* /etc/libvirt/
/bin/rm /etc/libvirt/nwfilter/vdsm-no-mac-spoofing.xml
ln -s ../default.xml /etc/libvirt/qemu/networks/autostart
/bin/cp $CLEAN/logrotate.d/* /etc/logrotate.d/
/bin/rm /etc/ovirt-hosted-engine/answers.conf
rm -rf /etc/ovirt-hosted-engine/firewalld
/bin/cp $CLEAN/ovirt-hosted-engine-ha/* /etc/ovirt-hosted-engine-ha/
/bin/cp $CLEAN/sysconfig/* /etc/sysconfig/
/bin/rm /etc/vdsm/vdsm.conf
/bin/rm /etc/pki/vdsm/*/*.pem
/bin/rm /etc/pki/CA/cacert.pem
/bin/rm /etc/pki/libvirt/*.pem
/bin/rm /etc/pki/libvirt/private/*.pem

chkconfig --levels 345 libvirtd on
=========================================================================
And: Deleted on the nfs server the content of the data domain export.

This does not revert the changes to network interfaces (e.g. addition of
ovirtmgmt bridge).

Best regards,
-- 
Didi