On Mon, Apr 17, 2017 at 4:03 PM, martin chamambo <chamambom@gmail.com> wrote:
I had issues with my master data storage domain and the only way was to set the hosts in maintenence mode ,delete the datacenter so i can recreate it....

I managed to delete the datacenter and the storage domains , and created a new datacenter , but now the existing clusters are not part of a datacenter and trying to add them to the datacenter gives out this error below

Error while executing action: Cannot edit Cluster. Changing management network in a non-empty cluster is not allowed.

and by non empty i guess it means there is a host inside that cluster ? fair enough

trying to change a node cluster comes with the Host Cluster Dropdown with this

Datacenter:Undefined

How can i fix the catch 22 scenario without deleting clusters or hosts..


surely there should be a smarter way ?


I don't know how to manage your current situation.
But I somehow had a similar problem as your initial one a few weeks ago but for other reasons.
My case was that I defined an iSCSI DC and then used a LUN to create an iSCSI storage domain.
I initially formatted and tested the infrastructure to manage all the relevant configuration (iSCSI config, multipath config, ecc.) before going to production.
Then I had to decommission/re-create this storage domain at its target stage: my iSCSI lun had to be a raw copy of a SAN FC lun I migrated from older DC.
But in the mean time I created a cluster of two hosts and such.... and I discovered that I was in problem as you described....
My solution was to create a smaller NFS share and add it as a new storage domain: in old releases of oVirt it was not possible to mix different types of storage domains in the same DC, but since 3.4 (oVirt version and DC version) it is:
http://www.ovirt.org/develop/release-management/features/storage/mixed-types-data-center/

With this workaround I was able to put then my iSCSI storage domain into maintenance and have the small NFS one to become the master.
And finally to remove the iSCSI storage domain and import the copied one, finally decommissioning the temporary NFS share.

Not the best way but it could have helped you too, eventually setting a directory of the host itself as a share for a temporary operation of this kind.
In the past it was discussed about SPM removal and by consequence the removal of master storage domain concept... but probably it didn't get any update.
See this whole thread for example:
http://lists.ovirt.org/pipermail/users/2016-May/039782.html

For sure it would be nice to have a clean way (if we are not missing some other correct workflow) to manage the case when you have only one SD and for some reason you need to scratch it but preserve your DC/cluster config.
HIH,
Gianluca