It seems you are using a custom build of Flume - the type for your sink is
not a FQCN or one of the defined types in SinkType.
It would be helpful if you inform us with the following additional
1. The Flume version you updated for this custom build.
2. The log files in both agents when this error occurred.
This will help out in the diagnosis of the root cause.
If I were to take a wild guess, there are possible errors taking place in
one or both agents.
Also, you are using MemoryChannel which is volatile and non-persistent.
Stoping the agent effectively kills the JVM process and discards any events
that may have been buffered the process memory for this channel.
Any events buffered in the memory channel before the agent was stopped will
be lost permanently.
I personally only use memory channels when doing small tests. I dont use it
in production events where I cannot afford to lose the events captured by
I would recommend that you send us your log files, including the version of
Flume NG with a summary of the channels you made for your custom build.
Also, try changing the channel type to FileChannel and see if you still
have this issue when you restart the agents.
On 12 May 2013 21:35, cherubimsun <[EMAIL PROTECTED]> wrote:
> *hi flumes,*
> Can somebody help me ASAP?
> When i have restart the second layer agent,the the first layer agent can
> not send the the log to the second's layer and no error print out.
> Here is the environment(the log has been send from first layer
> agent(t188) to second agent(t185)):
> agent_1st[192.168.11.188 agent] ---->agent_2nd[192.168.11.185 agent]
> After i'm fllowing the steps,the flume ng is not working:
> 1.start the agent_2nd(t185) and agent_1st(t188).
> 2.send the log.(fond the log has been sended to the agent_2nd).
> 3.stop the agent_2nd.
> 4.delete the log.
> 5.start the agent_2nd.
> 6.send the log from agent_1st 2 agent_2nd.(There is nothing in the log
> folder. And the flume has't error info).
> Thanks a lot!
*American Institute for Computing Education and Research Inc*