Hi Roshan,
By destination server i meen the flume in front of our hadoop cluster.
eg: the last component in the following flow :
source-server -> tcp/syslog -> flume-spool (memory + spooling) -> avro
-> flume-destination (hadoop-cluster)
I didn't restart anything, i just stopped the last flume (hadoop).
Then the flume-spool should continue to spool in its memory/disks but it
simply stopped receiving.
Regards,
--
Smaine Kahlouch - Engineer, Research & Engineering
Arkena | T: +33 1 5868 6196
27 Blvd Hippolyte Marquès, 94200 Ivry-sur-Seine, France
arkena.com
On 03/23/15 22:55, Roshan Naik wrote:
Catching up on emails post vacation…
- By "destination server" Do you mean the 2nd flume agent in the flow ?
- The NullPointer is probably indicating a bug somewhere. Your logs
seem to indicate there a restart of the agent with the spillable
channel. Can you confirm if this is explicit restart or was it an
automatic dynamic restart due to config file update ?
-roshan
From: Smaine Kahlouch <[email protected]
<mailto:[email protected]>>
Reply-To: "[email protected] <mailto:[email protected]>"
<[email protected] <mailto:[email protected]>>
Date: Thursday, March 5, 2015 3:53 AM
To: "[email protected] <mailto:[email protected]>"
<[email protected] <mailto:[email protected]>>
Subject: Spillable memory perf tests : source stops sending
Hello,
I'm currently testing the spillable memory channel.
The idea is to have a fast channel coupled with a file channel in
order to spool logs when our hadoop cluster is unreachable.
To give you the whole architecture :
source-server -> tcp/syslog -> flume-spool (memory + spooling) -> avro
-> flume-destination (hadoop-cluster)
My test scenario is :
- send 100 000 logs from the source server
- stop the flume-destination during this sending
- wait a few seconds in order to spool on the flume-spool server
- start the flume-destination
- check if all the logs have been received on the flume-destination
server.
The problem is that when i stop the destination server, the source
server stops sending *completely*.
Is this the normal behaviour ? Should it spool on the flume server
instead of stopping the source ?
On my flume logs i can see the following exception :
05 Mar 2015 10:22:13,828 INFO
[SinkRunner-PollingRunner-DefaultSinkProcessor]
(org.apache.flume.sink.AbstractRpcSink.createConnection:217) - Rpc
sink sin1: Building RpcClient with hostname: 10.115.77.6, port: 55555
05 Mar 2015 10:22:13,828 INFO
[SinkRunner-PollingRunner-DefaultSinkProcessor]
(org.apache.flume.sink.AvroSink.initializeRpcClient:126) - Attempting
to create Avro Rpc client.
05 Mar 2015 10:22:13,862 WARN
[SinkRunner-PollingRunner-DefaultSinkProcessor]
(org.apache.flume.api.NettyAvroRpcClient.configure:634) - Using
default maxIOWorkers
05 Mar 2015 10:22:13,869 ERROR
[SinkRunner-PollingRunner-DefaultSinkProcessor]
(org.apache.flume.SinkRunner$PollingRunner.run:160) - Unable to
deliver event. Exception follows.
java.lang.NullPointerException
at
org.apache.flume.channel.SpillableMemoryChannel$SpillableMemoryTransaction.doRollback(SpillableMemoryChannel.java:590)
at
org.apache.flume.channel.BasicTransactionSemantics.rollback(BasicTransactionSemantics.java:168)
at
org.apache.flume.sink.AbstractRpcSink.process(AbstractRpcSink.java:394)
at
org.apache.flume.sink.DefaultSinkProcessor.process(DefaultSinkProcessor.java:68)
at org.apache.flume.SinkRunner$PollingRunner.run(SinkRunner.java:147)
at java.lang.Thread.run(Thread.java:745)
Please find enclosed the flume configurations (spooling server and
destination server).
Regards,
--
Smaine Kahlouch - Engineer, Research & Engineering
Arkena | T: +33 1 5868 6196
27 Blvd Hippolyte Marquès, 94200 Ivry-sur-Seine, France
arkena.com
