When a large view is requested (many data) or the database is requested to build all views and compact at the same time.
The beam.smp couchdb process tries to start a new couchjs process which does not get any main memory.
- Open the couchdb configuration
- set in Section 'query_server_config' the 'os_process_limit' to '1' (this is the maximum number I figured out for an aws micro instance)
- If you wish to replicate (continous) between two or more couchdb-instances, you'll need to set the 'os_process_limit' to 2. Otherwise the replication takes this process and indexes and so on will never be updatet.