Host Manager (hm)
Syntax
smash hm [-megactrl multicastControlAddress] [-scriptdir scriptDir] [-path execPath] [-conffile megaConfFile]
smash hm [-megactrl multicastControlAddress] [-scriptdir scriptDir] [-path execPath] [-conffile megaConfFile] [-fork] [-target numHostManagers] [-int checkInterval] [-path execPath] [-rcmd execCmd] [-log logFile] [-nolog] [-glunix] [-lb]
Description
The Host Manager (hm) is the key component of the AS1 framework. A host manager listens on the multicast control address (-megactrl) for service requests, and launches servents to answer these requests. There are two major ways to run hm, either standalone, or as part of a pool of host managers. To run a pool of host managers, the -fork, -target and -path options are used, and the target number of hm's will be launched on machines in the pool. In standalone mode, no command line options are required, and only a single hm is run.
Command Line Arguments
- -megactrl multicastControlAddress
multicastControlAddress is a shared multicast channel which all host managers, clients and servents use to announce and listen. This defaults to 224.4.5.24/50000/31.
- -scriptdir scriptDir
The directory where servent scripts are placed after being downloaded from the web. Defaults to /var/tmp/.
- -path execPath
The path to use when executing servent or hm scripts. If the fork option is being used, this path must be valid across all hm host machines.
- -conffile megaConfFile
The location of the mega/AS1 configuration file. This file is not required for operation, but is used to specify the scopes for setting up "mega chains".
Pool Command Line Arguments
- -fork
If this option is set, the hm will periodically announce its presence on the shared channel. If it hears that the number of hms announcing on the channel is less than the target number (below), it will attempt to fork a new hm on one of the machines listed in the file "./hmhosts". The hmhosts file is required.
- -target numHostManagers
This sets the target number of host managers which should be running at any given time. The group of hms will collectively fork new hm's and commit suicide in order to reach this number.
- -int checkInterval
This is the interval (in ms) in which the host manager spawning algorithm operates.
- -rcmd execCmd
Denotes the exec command used to fork hm's on other machines. Defaults to "ssh -n".
- -log logFile
The directory and filename to use for keeping the hm log. Defaults to "/tmp/hmlog". Log is only kept when hm is part of a pool. Otherwise output goes to the command line.
- -nolog
Indicates that no log should be kept.
- -glunix
Berkeley-specific option to indicate that the hm pool is running on the Berkeley NOW cluster and can use glunix-specific commands.
- -lb
Perform load balancing.