On Sat, Feb 24, 2018 at 2:32 PM, Alessandro De Salvo <Alessandro.DeSalvo@roma1.infn.it> wrote:
Hi,

I have just migrated my dev cluster to the latest master, reinstalling the engine VM and reimporting from a previous backup. I'm trying with 4.3.0-0.0.master.20180222192611.git01e6ace.el7.centos

I had a few problems:

- the documentation seems to be outdated, and I just find by searching the archives that it's needed to add the two (undocumented) options --he-remove-storage-vm --he-remove-hosts

- despite the fact I selected "No" to running the engine-setup command in the VM (the ovirt appliance), the engine-setup is executed when running hosted-engine --deploy, and as a result the procedure does not stop allowing to reload the db backup. The only way I found was to put the hosted-engine in global maintenance mode, stop the ovirt-engine, do an engine-cleanup and reload the db, then it's possible to add the first host in the GUI, but must be done manually

- after it's all done, I can see the hosted_storage is imported, but the HostedEngine is not imported, and in the Events I see messages like this:

VDSM atlas-svc-18 command GetVolumeInfoVDS failed: Image path does not exist or cannot be accessed/created: (u'/rhev/data-center/mnt/glusterSD/atlas-fsserv-07.roma1.infn.it:_atlas-engine-02/f02d7d5d-1459-48b8-bf27-4225cdfdce23/images/c815ec3f-6e31-4b08-81be-e515e803edce',)

   the path here is clearly wrong, it should be /rhev/data-center/mnt/glusterSD/atlas-fsserv-07.roma1.infn.it:_atlas-engine-02/f02d7d5d-1459-48b8-bf27-4225cdfdce23/images/b7bc6468-438c-47e7-b7a4-7ed06b786da0/c815ec3f-6e31-4b08-81be-e515e803edce, and I see the hosted_engine.conf in the shared storage has it correctly set as vm_disk_id=b7bc6468-438c-47e7-b7a4-7ed06b786da0.


Any hint on what is not allowing the HostedEngine to be imported? I didn't find a way to add other hosted engine nodes if the HE VM is not imported in the cluster, like we were used in the past with the CLI using hosted-engine --deploy on multiple hosts.

Ciao Alessandro,
with 4.2.1 we introduced a new deployment flow for hosted-engine based on ansible.
In this new flow we run a local VM with a running engine and we use that engine to create a storage domain and a VM there.
At the end we shutdown the locally running engine and we move it's disk over the disk of the VM created by the engine on the shared storage. At this point we don't need anymore the autoimport process since the engine migrated there already contains the engine VM and its storage domain.

We have an RFE, for this new flow, to add a mechanism to inject an existing engine backup to be automatically restored before executing engine-setup for migration/disaster-recovery scenarios.
Unfortunately it's still not ready but we have an hook mechanism to have hosted-engine-setup executing custom ansible tasks before running engine setup; we have an example in /usr/share/ovirt-hosted-engine-setup/ansible/hooks/enginevm_before_engine_setup/enginevm_before_engine_setup.yml.example

Otherwise the old flow is still there, you have just to add --noansible and everything should work as in the past.

 

Thanks for any help,


    Alessandro

_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users