Nikola Mandic created FLUME-2659:
------------------------------------
Summary: s3 sink unable to recover
Key: FLUME-2659
URL: https://issues.apache.org/jira/browse/FLUME-2659
Project: Flume
Issue Type: Bug
Components: Sinks+Sources
Affects Versions: v1.5.2
Environment: linux running on aws with hdfs library
https://github.com/twitter/hadoop-lzo.git
https://archive.apache.org/dist/hadoop/core/hadoop-2.4.0/hadoop-2.4.0.tar.gz
this is the docker file
````
RUN apt-get install -y wget default-jre git make curl vim tmux make gcc
autoconf build-essential libc-dev maven liblzo2-2 liblzo2-dev default-jdk
RUN cd ~
RUN wget
http://ftp.halifax.rwth-aachen.de/apache/flume/1.5.2/apache-flume-1.5.2-bin.tar.gz
RUN tar xf apache-flume-1.5.2-bin.tar.gz
ENV HOME /root
WORKDIR /root/apache-flume-1.5.2-bin
RUN rm -r /root/apache-flume-1.5.2-bin/conf
ADD ./conf /root/apache-flume-1.5.2-bin/conf
WORKDIR /root/
RUN git clone https://github.com/twitter/hadoop-lzo.git
RUN wget
https://archive.apache.org/dist/hadoop/core/hadoop-2.4.0/hadoop-2.4.0.tar.gz
RUN tar xf hadoop-2.4.0.tar.gz
WORKDIR /root/hadoop-2.4.0
RUN rm -r /root/hadoop-2.4.0/etc
ADD ./etc /root/hadoop-2.4.0/etc
WORKDIR /root/hadoop-lzo
ENV JAVA_HOME /usr/lib/jvm/java-7-openjdk-amd64
RUN C_INCLUDE_PATH=/usr/local/lzo-2.06/include
LIBRARY_PATH=/usr/local/lzo-2.06/lib mvn clean test
RUN mvn install
RUN cp target/hadoop-lzo-0.4.20-SNAPSHOT.jar ~/apache-flume-1.5.2-bin/lib/
ENV CLASSPATH
/usr/lib/jvm/java-7-openjdk-amd64/lib/:/root/apache-flume-1.5.2-bin/lib/*:/root/hadoop-2.4.0/share/hadoop/tools/lib/*:/root/.m2/repository/org/apache/hadoop/hadoop-common/2.4.0/hadoop-common-2.4.0.jar:/root/hadoop-2.4.0/etc/hadoop/
````
Reporter: Nikola Mandic
I use hdfs sink to write to s3.
Often s3 fails for some reason for example
```org.apache.flume.sink.hdfs.HDFSEventSink.process:463) - HDFS IO error```.
Sometimes it can be something else like reporting that it can not continue from
bad mark etc.
The problem is that hdfs sink does not recover unless I do manual restart.
What to do in those situations when s3 fails after running for long long time?
It would be quite nice if hdfs sink would restart connection completely instead
of me.
Right now it just bumps into error from time to time and stops working until
restart.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
