Database Reference
In-Depth Information
return
status;
}
Sink Processors
Sink processors are a mechanism for controlling the parallelism of a
channel. Essentially, Sink processors are used to define groups of sinks that
are then used for load balancing or, possibly, for failover. For streaming
applications, the failover application is not as useful as the load-balancing
application.
Toactivate load-balancing sinkprocessing onachannel, setthe
processor
.type
property of the sink to be
load_balance
and use the
processor.sinks
property to define the sinks that should be used for the
group:
agent_1.sinkgroups= group-1
agent_1.sinkgroups.group-1.sinks= sink-1 sink-2
agent_1.sinkgroups.group-1.processor.type= load_balance
agent_1.sinkgroups.group-1.processor.selector= random
After this definition, the channel is assigned to the group rather than the
sink, and events will be randomly sent to
sink-1
or
sink-2
. The
load-balancing processor can also use a round-robin selection method by
changing the selector to
round_robin
instead of
random
.
Flume Channels
Flume channels are the temporary storage mechanism between a source
and a sink. There are three different types of channel used in production
Flume environments: memory channels, file channels and JDBC channels.
These channels are listed in terms of increasing safety and decreasing
performance.
Memory Channel
The Memory channel is the fastest of the available Flume channels. It is also
the least safe because it stores its events in memory until a sink retrieves
them.Iftheprocessgoesdownforanyreason,anyeventsleftintheMemory
channel are lost.
