Open rblaine95 opened 7 years ago
java-installer-a4d52ba0.sh
has an error in the find_java
method --
find_java() {
if [[ -z "$JAVA_HOME" ]]; then
export JAVA_HOME=$(find $1 -name $2 | head -n1)
fi
}
The find returns 2 entries:
/usr/lib/jvm/java-8-openjdk-amd64
and /usr/share/gdb/auto-load/usr/lib/jvm/java-8-openjdk-amd64
.
The order of these entries differs between machines (For some reason, when I deploy this on a locally hosted machine the /usr/share
entry is first whereas the /usr/lib
entry is second. This causes java to break).
To Fix:
Pipe the output of the find through a sort
. This will ensure that /usr/lib
will always be at the top.
Otherwise set $1
to /usr/lib
instead of /usr/
.
find_java() {
if [[ -z "$JAVA_HOME" ]]; then
export JAVA_HOME=$(find $1 -name $2 | sort | head -n1)
fi
}
Also please upgrade this to Xenial (and please update the Zeppelin mirror. Zeppelin 0.5.2 is way too out of date. We need 0.7.2). I have been fighting with the Apache Bigtop Spark Processing deployment for several days because I need an R interpreter and, for some idiotic reason, the official Apache Bigtop Zeppelin JuJu Charm doesn't support R in any straightforward way.
To update the Zeppelin mirror, modify resources.yaml
and replace the current Zeppelin entries with:
"zeppelin-x86_64":
"url": "http://www-eu.apache.org/dist/zeppelin/zeppelin-0.7.2/zeppelin-0.7.2-bin-all.tgz"
"hash": "068acacdf00321ddb298cb31b9e676be2e72f2de985e78a95d239bdcf37a85037eacdb24b00b692f5cefe904f18f624648e9eb6cc4cfff67425c442e2a2caa7c"
"hash_type": "sha512"
When deploying this bundle on a localhost LXD platform, the hadoop units try to install
openjdk-8-jdk
and continuously fails.Deploying the same bundle on an Azure platform has no errors and installs
openjdk-7-jdk
instead.