This page will try to highlight different features and their maturity based on stories gathered from the community.
States are
- Recommended - feature is known to be in daily use in multiple places, and recommended for use. The API is stable.
- Beta - feature is usable, known to be in use at one or to places, but may have some issues or limitations. The API is close to stable.
- Experimental/Development - feature is in development and will likely have issues. API may change significantly. Be prepared to go into code.
- Testing - for debugging and testing purposes only.
v0.9.4 (pre-apache release)
Flume Master features
Feature |
Status |
Notes |
---|---|---|
Master Mode |
|
|
Single Master ZK-backed mode |
|
Default. Recommended mode of use |
Single Master Memory-backed mode |
|
Useful for development and debugging |
Multi-Master ZK-backed mode |
|
This is being used by a handful of users but has known limitations. (e.g. no support for auto-agent collector mapping |
Automatic configurations |
|
|
Automatic agent-collector mapping |
|
Must use Single Master mode. All logical nodes must be up before this feature will work. May have problems with reconfiguration or if configuration written before logical nodes report |
Automatic flow isolation |
|
Must use Single Master mode. Dependent upon Automatic agent-collector mapping |
Metrics |
|
|
Master JSON Metrics |
|
Looking for feedback and nagios/munin/etc integration stories. |
Flume Node features
Feature |
Status |
Notes |
---|---|---|
Agents and Collectors |
|
|
agentBESink |
|
|
agentDFOSink |
|
|
agentE2ESink / agentSink |
|
|
collectorSource |
|
|
collectorSink("hdfs://...",..) |
|
HDFS is the recommended target file system. |
collectorSink("s3n://...",..) |
* |
In use but has known issue due to how s3 files are "closed" |
collectorSink("file:///...",..) |
|
Intended for testing but could be used in production |
agentBEChain |
|
Need known production use stories |
agentDFOChain |
|
Need known production use stories |
agentE2EChain |
|
Need known production use stories |
auto*Chain |
|
Limitation on at the master, to be address in future versions |
Sources |
|
|
thriftSource |
|
Default RPC. Recommended mode of use |
avroSource |
|
Need known production use stories |
syslogTCP source |
|
|
tail/multiTail/tailDir |
* |
This is known to have some duplication issues but has been used in production settings. If this is encountered, known workarounds include using exec source, or having applications write to flume via RPC |
exec source |
|
|
log4j appender sources |
|
|
text |
|
|
Sinks |
|
|
thriftSink |
|
Default RPC. Recommended mode of use |
avroSink |
|
Need known production use stories |
attr2hbase, hbase sink plugin |
|
Known to be in use in a few places |
seqfile sink |
|
Used internally (E2E mode) |
dfs sink |
|
Used internally |
Metrics |
|
|
Node JSON metrics |
|
Looking for feedback and nagios/munin/etc integration stories. This api is likely to change as we receive feedback |