tr6a-glossary.asciidoc

Glossary

• secondarynn (aka "secondary namenode") — handles compaction of namenode directory. It is NOT a backup for the namenode.

• support daemon  — one of namenode, datanode, jobtracker, tasktracker or secondarynn.

• job process (aka "child process")  — the actual process that executes your code

• tasktracker  — interface between jobracker and task processes. It does NOT execute your code, and typically requires a minimal amount of RAM.

• shuffle merge  — (aka shuffle and sort)

• shuffle buffer  — 

• map sort buffer  — 

• Resident memory (RSS or RES)  — 

• JVM  — 

• JVM heap size  — 

• Old-Gen heap  — 

• New-Gen heap  — portion of JVM ram used for short-lived objects. If too small, transient objects will go into the old-gen, causing fragmentation and an eventual STW garbage collection.

• garbage collection  — 

• STW garbage collection  — "Stop-the-world" garbage collection, a signal that there has been significant fragmentation or heap pressure. Not Good.

• attempt ID  — 

• task ID  — 

• job ID  — 

Questions:

• "task process" or "job process" for child process? *