I don't think there's a combination of options you can set to get the desired behaviour out of the box. I think the script evolved into this rather than was designed - the CDH component downloading was added into the existing script after that option was added.
You can work around this by changing IMPALA_TOOLCHAIN. E.g. running IMPALA_TOOLCHAIN=/tmp/toolchain ./bin/bootstrap_toolchain.py to download the toolchain to /tmp. What I essentially did is did the initial bootstrap, then set SKIP_TOOLCHAIN_BOOTSTRAP=true and then symlink custom-build toolchain components into the toolchain dir. It would probably make sense to add some command line flags to bootstrap_toolchain.py to control what it downloads since I'm sure other people will run into this. On Thu, Sep 29, 2016 at 2:56 AM, Lars Volker <[email protected]> wrote: > Hi, > > I tried to follow the instructions here: > https://cwiki.apache.org/confluence/display/IMPALA/ > Building+native-toolchain+from+scratch+and+using+with+Impala > > However I keep running into the following error: > > SKIP_TOOLCHAIN_BOOTSTRAP is true, skipping download of Python dependencies. > SKIP_TOOLCHAIN_BOOTSTRAP is true, skipping toolchain bootstrap. > cp: target > ‘/home/lv/i4/toolchain/cdh_components/hadoop-2.6.0-cdh5. > 10.0-SNAPSHOT//lib/native’ > is not a directory > Error in ./buildall.sh at line 281: cp > "$HADOOP_LZO"/build/native/Linux-*-*/lib/libgplcompression.* > "$HADOOP_HOME/lib/native" > > The problem seems to be that CDH component download is disabled by > SKIP_TOOLCHAIN_DOWNLOAD, too. > > Is there a way around this? > > Thanks, Lars >
