Skip to content
Vicente Adolfo Bolea Sanchez edited this page Nov 16, 2017 · 4 revisions

How to configure VeloxDFS

Installing VeloxDFS is quite a challenge, yet configure it is not far from easy.

VeloxDFS relies on configuration files ".eclipse.json" to obtains its setups. There are several variables that need to be used in order to make Velox work.

Where to write .eclipse.json file

It depends whether you want to have a system wide configuration file or not. Velox will give priority to the ~/.eclipse.json than to the system wide /etc/eclipse.json.

Essential variables and their meanings:

Network settings

"network" : {
"ports" : {
  "internal" : 8318,             # Port used internally in Velox
  "client"   : 8319,             # Port used between client and Velox
  "mapreduce"   : 8321           # Port used for Mapreduce engine (Only for VeloxMR) 
},

"iface"          : "eth0",       # Interface name
"nodes"          : [             # List of nodes to run VeloxDFS
  "172.20.1.40",
  "172.20.1.41",
  ...
  "172.20.1.47",
  "172.20.1.48",
  "172.20.1.49"
],
"serialization"   : "binary"    # Either binary or xml (use xml for debugging)
},

File system settings

"filesystem" : {
  "block"  : 2000,      # Size of block (in Bytes)
  "buffer" : 512,       # Size of file buffers (in Bytes)
  "replica" : 3         # Number of replicas per block
},

"path" : {
  "scratch"      : "/scratch/vicente/storage",   # Where to save blocks
  "metadata"     : "/scratch/vicente/"           # Where to save metadata file
},

Logger settings

"log" : {
  "type" : "LOG_LOCAL6",    # Log type for rsyslog (One per user)
  "name" : "ECLIPSE",       
  "mask" : "info"           # Log messages above this log level
},