Best-practices sharding complete MongoDB -

we changed logic in our mongodb-setup follows: half of replica-sets in cluster contain databases , corresponding collections not sharded. other half contain sharded collections.

while databases still spread around replica-sets curious know: there disadvantage in having database on server doesn't hold of collections? there kind of meta-data held mongodb waste ressources?

is there disadvantage in having database on server doesn't hold of collections?

this never case sharding. collections start on first (master) server, contains database (i think of enabling sharding on database level, do), , scale out across other shards (replica sets).

the databases , collections made on other shards data needed moved them.

as such further questions have been mad inert knowledge.

edit

"which contains database"

i mean copy of database.

this hinted next paragraph state databases , collections made on each shard needed.

the shard, or rather replica set in case, ever know own data , own data ever exist when data required.

the config servers ones map shards range boundaries.

i suppose, think it, if have collection on 1st node not on nth node while have collection there thinking house unneeded metadata. however, still not, not until collection made , added namespace files, @ point data there.