Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On 12/08/2016 08:05 AM, Stephane Gaucher wrote: > I would like to move my cluster project from pre-production to > production mode. To do this I absolutely must be able to restart this > cluster automatically and quickly. Currently, both nodes are virtual > Ubuntu 16.04 servers running virtualbox. Periodically my backup > procedure exports to an appliance (.ova file) the servers. However, when > importing this appliance, partition synchronization no longer works. The > cluster runs but is out of sync. Server installation is standard and > synchronization takes place on a second virtual disk. DRBD synchronizes data based on generation ID's stored in each devices metadata. By restoring from an exported ova, you're essentially reverting the metadata to the point in time you took the backup. To DRBD this export/import will either look like a split-brain, or a completely unrelated device. For more info on generation IDs and how DRBD decides which direction to sync data, see: https://www.drbd.org/en/doc/users-guide-84/s-gi It might be best to export your OVA after having created new metadata for your DRBD devices. That way the generation IDs are zeroed, and when the newly restored node rejoins the cluster, it will sync from the other node without needing human interaction. Otherwise, you just need to resolve the split brain, or recreate metadata, to get DRBD syncing again. Hope that helps. -- Matt Kereczman – Cluster Engineer +1 503 573-1262 x205 matt at linbit.com LINBIT | Keeping the Digital World Running DRBD – Corosync – Pacemaker DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.