Currently, if OSM master crashes and we either replace it or switch one of slaves in its place, OSM data has to be reimported to make replication work. To make things easier, the following files need to be backed up/replicated to other servers:
- nodes.bin - flat nodes file
- state.txt - replication state
- osm_expire/* - expired nodes list, in case some changes weren't fed to Tilerator yet
- Some future Tilerator's osm_expire ingestion state (TBD).
What are our options, rsync?