[OSM-dev] rsync and osm2pgsql flat nodes

[Paul Norman]

osm2pgsql flat nodes is a binary file with the position of all nodes, in 
a representation which is very efficient for full planet dumps, needing 
only 8 bytes * max node ID, instead of the 40-60 bytes * number of nodes 
that alternative representations would need. For the planet where the 
number of nodes is about 81% of the max node ID, this takes much less 
space.

One significant downside for production environments is that this data 
is not in the database and does not have the replication and redundancy 
of in-DB nodes. This is a problem but should not be insurmountable. 
Rather than needing to keep the flat nodes and database synchronized 
with your replication method, the flat nodes needs to be no older than 
the state file which also should be saved.

Has anyone used rsync to keep a copy of the flat nodes file up to date 
on another standby machine which can then step in in case the first one 
fails? My initial thought was to have another cron script that runs 
async to the updates which rsyncs the flat nodes and then the state 
file, in that order. I have no idea how the performance is, but most of 
the flat nodes file should not have changed.