T128295 was annoying!
We don't know why the JVM processes locale was messed up, but if we manually set file.encoding for all JVM processes, this wouldn't happen. Ideally we could set this for all JVMs, but we might have to manually set it in all of the individual JVM CLIs.
| Project | Branch | Lines +/- | Subject | |
|---|---|---|---|---|
| operations/puppet | production | +7 -0 | Set file.encoding=UTF-8 for all java processes in analytics cluster |
According to: http://javarevisited.blogspot.com/2012/01/get-set-default-character-encoding.html it seems it's possible to set an environment variable:
export JAVA_TOOL_OPTIONS=-Dfile.encoding="UTF-8"
And the jvm should pick that up on each run. I tested this on my local machine and it worked.
Change 276010 had a related patch set uploaded (by Ottomata):
Set file.encoding=UTF-8 for all java processes in analytics cluster
Change 276010 merged by Ottomata:
Set file.encoding=UTF-8 for all java processes in analytics cluster