Essex and Folsom for Multi-storage Drivers

[GERNETAPP]

Hi,

   We are planning a migration from Essex (actual production environment) to Folsom and then from Folsom to Grizzly, but the main issue comes with the first step in the migration process. We actually had a DFM server that it's configured with 3 ResourcePools (each with an Aggregate of a different Filer), all of those Filers are 7-Mode. In the cinder.conf file we had the netapp 7-mode driver along with the DFM server IP address and all the corresponding parameters. Recently we bought a Cluster-mode Filers, and we need to add these Cluster-mode filers to the existing Essex instalation so we could mapped to Essex, and also to Folsom, so in that way we could migrate all the volumes and the virtual machines. The thing is that for the cluster-mode to run correctly we need to change the driver on the cinder.conf file, but we cannot delete the old 7-mode driver. It's possible or someone done this kind of migration? It's possible to do it in another way? or someone has the driver hacked to support multi-drivers in Cinder on Essex/Folsom? One thing to point is that the 7-Mode filers all of them had only the vFiler0, so no of them had vFilers created.

Thanks a lot,

Best regards,

[bswartz · Accepted solution]

You're going to want to create a new volume on the cluster for each volume that exists on the 7-mode side, and dd the data from the old volume to the new volume (OpenStack volume not NetApp volume). There is some built in support for this kind of operation coming in OpenStack Havana, but as of now, it will be a pretty manual operation. Still, it would not be hard to script it. The flow would look like this:

for each instance

     (optional) shut down the instance

     for each volume attached to the instance

          detach volume from instance

          create new volume of equal size

          attach old and new volume to this machine

          dd the data

          detach the volumes from this machine

          attach the new volume to the instance

          destroy the old volume (optional)

     end for

     start the instance (if we shut it down above)

end for

for each volume that's unattached

     create new volume of equal size

     attach old and new volume to this machine

     dd the data

     detach the volumes from this machine

     destroy the old volume (optional)

end of