-
Type: Bug
-
Resolution: Done
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
Labels:None
Considering the scale of data most people considering the "Lots of Collections" route have, I feel the page could use a lot more detailed information about the ramifications of choosing to use lots of collections.
The following is what I added to my internal wiki:
=== Namespaces ===
- namespaces = collections + indexes and are per-database (stored in a db.ns file)
- Check namespaces with ''db.system.namespaces.count()'' (safe & cheap)
- ''show collections'' in the mongo shell will no longer work
- ~600 bytes per namespace on disk
- namespaces are stored in a hashtable
- so you should only fill the namespace table up to ~75%
- namespace database is statically sized on server startup
- change with --nssize
- defaults to 16mb
- can go up to 2047mb (docs say 2GB, but --nssize=2048 crashes)
- db.ns file is created and zeroed to nssize //the first time data is inserted into the database//
- So we'd need to be careful creating test dbs on busy servers with high nssizes
== Limits ==
Limits are approximate due to the fact hash tables should never be full.
- Default: *20k*
- Maximum: *2.5m*
- duplicates
-
DOCS-661 Create Data Modeling Page
- Closed