We just hit a stupid problem on our elasticsearch installation with
our self written river.
The river is quite similar to the couchdb one, connecting to a host
and getting some data
I am rereading the configuration on every call of the slurper thread,
as it can be disabled at runtime or the interval to query the host
should be changed, someone possibly might misconfigure it, resulting
in some exception, for example a NumberFormatException.
However runtime exceptions in rivers seem not to be caught in any way,
resulting in dying of the thread without any notice. Is this intended?
Apart from that the river functionality is dead cool, importing data
until it can be found is now a matter of seconds in our platform.