dmlc
xgboost is compiled in local mode
Hi,
really struggling to compile xgboost for yarn with hdfs support. I can launch the yarn app but in the container logs it show:
terminate called after throwing an instance of 'std::runtime_error'
what(): xgboost is compiled in local mode
to use hdfs, s3 or distributed version, compile with make dmlc=1
The libhdfs library on my system is placed as follows:
/usr/lib64/libhdfs.so
/usr/lib64/libhdfs.so.0.0.0
/usr/lib/hadoop/lib/native/libhdfs.a
/usr/include/hdfs.h
I tried building xgboost using these options:
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib64 JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64 HADOOP_HDFS_HOME=/usr
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib/hadoop/lib/native/ JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64 HADOOP_HDFS_HOME=/usr
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib64 HDFS_INC_PATH=/usr/include JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib/hadoop/lib/native/ HDFS_INC_PATH=/usr/include JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64
But still no joy, keep getting the "xgboost is compiled in local mode".
Can you please advise?
Thanks, Antony.
ok, I managed to build it using:
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib64 HDFS_INC_PATH=/usr/include JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64 dmlc=../dmlc-core/
however that makes it dependent on libglog so I had to hack repo/xgboost/Makefile:
@@ -2,7 +2,7 @@ export CC = $(if $(shell which gcc-5),gcc-5,gcc)
export CXX = $(if $(shell which g++-5),g++-5,g++)
export MPICXX = mpicxx
-export LDFLAGS= -pthread -lm
+export LDFLAGS= -pthread -lm -lglog -L../../deps/lib
This however now makes the yarn app dependent on lobglog.so I have to submit dmlc_yarn.py with -u deps/lib/libglog.so.0 which is bit dirty.
so final method to get it all working without the libglog dependency:
cd repo/xgboost
git clone https://github.com/dmlc/dmlc-core
cd dmlc-core
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib64 HDFS_INC_PATH=/usr/include JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64
cd ..
make USE_HDFS=1 HDFS_LIB_PATH=/usr/lib64 HDFS_INC_PATH=/usr/include JAVA_HOME=/usr/lib/jvm/java-1.8.0-openjdk-1.8.0.45-28.b13.el6_6.x86_64 dmlc=dmlc-core
cp xgboost ../../bin/xgboost.dmlc
this is rather a hack/workaround, it is not straightforward build process given the declared wormhole build config options.
I'm having a similar issue, the tricks to tweak the build helped, however now I'm getting error like these:
Container: container_1447066627326_186565_01_000004 on s40819.dc4.local_8041
==============================================================================
LogType: stderr
LogLength: 130
Log Contents:
./xgboost.dmlc: error while loading shared libraries: libhdfs.so.0.0.0: cannot open shared object file: No such file or directory
LogType: stdout
LogLength: 0
Log Contents:
Well, the quick fix is to add libhdfs.so.0.0.0 using the -u option in dmlc_yarn.py .
Now my logs are clean
Container: container_1447066627326_186575_01_000028 on s40834.dc4.local_8041
==============================================================================
LogType: stderr
LogLength: 0
Log Contents:
LogType: stdout
LogLength: 0
Log Contents:
but the application still fails! I've attached the logs I've got from yarn logs -applicationId
Now I'm lost, hard to guess where to go from here. My guess is, something is wrong that I had to upload the libhdfs.so.0.0.0 file?
Any ideas?
