module Nn_config:Configuration of name nodessig
..end
class type nn_node_config =object
..end
val extract_node_config : Netplex_types.config_file -> nn_node_config
Nn_config.nn_node_config
for explanations):
netplex {
...
namenodes {
clustername = "cluster1"; (* required *)
node { addr = "hostname:port" };
node { addr = "hostname:port" };
...
hostname = "hostname"; (* optional *)
timeout = 60.0; (* optional *)
alive_min = 1; (* optional *)
alive_min_startup = 1; (* optional *)
elect_timeout 300.0; (* optional *)
rank = "0"; (* mandatory *)
replication = 3; (* optional *)
inodecache { port = <port> };
}
datanodes {
node { addr = "hostname:port" }
node { addr = "hostname:port" }
...
timeout = 60.0; (* optional *)
blocksize = 65536; (* mandatory *)
}
...
}
The parameter rank
can also be given by running a script that
outputs the rank to stdout. This script must be given as
parameter rank_script
.
Instead of describing the nodes directly in the config file
using node
subsections, one can also point to external file:
node_list = "filename";
port = <default_port>;
These two parameters can replace both the node
subsections for
namenodes
and for datanodes
. The external file enumerates the
hosts, one hostname a line (with shell-style comments).
Special values:
alive_min
and alive_min_startup
can also be set to "all"
(in double quotes). This means all configured name nodes must
be availablereplication
can also be set to "max" (in double quotes). This
means that the replication is set to the number of data nodes.val bmaprowsize : int