----------- | Whatsup | ----------- Description: ------------ Whatsup is cluster node up/down detection utility. The whatsup package includes: * libnodeupdown - A C library to determine up or down nodes in a cluster. See libnodeupdown(3) * Libnodeupdown.pm - A Perl module developed with Perl Extensions that takes advantage of the already available C library. Perl extensions is limited and provides an unorthodox Perl API. It is therefore recommended most users use Nodeupdown.pm. See Libnodeupdown(3). * Nodeupdown.pm - A Perl module that interfaces with Libnodeupdown.pm, hiding away the ugliness of Libnodeupdown.pm. See Nodeupdown(3). * whatsup - A command that outputs the up and down nodes in a cluster. See whatsup(1). Backend Tools: -------------- The nodeupdown library works in conjunction with one of several possible backend tools. The currently available backend tools are: Pingd: A simple daemon that regularly pings cluster nodes. Ganglia: See http://ganglia.sourceforge.net Cerebro: See http://cerebro.sourceforge.net The Pingd daemon is distributed along with whatsup as a subpackage. OpenIB: Query's the opensm (SA) for node up information. This requires the hostname to be reported as the HCA's "node description". Currently Mellanox HCA's and OFED 1.1 or greater provide for this information. More information can be found at http://openib.org Clusterlist Definitions: ------------------------ The nodeupdown library can also work in conjunction with several cluster databases. Genders - Uses a genders database and the genders library to determine all nodes in the cluster. (http://sourceforge.net/projects/genders) Hostsfile - Uses a file that lists the hostname of every node in the cluster on a separate line. Whatsup Options: ---------------- A few additional options may be available for the tool Whatsup dependent on your environment and the modules installed. The following are the currently supported options modules to Whatsup: cerebro-monitor: Supports the -M or --monitor option. Will monitor and output all changes in state to cluster nodes through the cerebro 'updown' event. Install: -------- For whatsup to work, you must install and configuration one of the backend tools for your system. The appropriate backend module for the libnodeupdown library must then be installed. If more than one module is installed, the first one discovered will be used. A clusterlist module and clusterlist database does not need to be installed for whatsup to work, however it would be beneficial to define one. Comments, patches, fixes, suggestions are always welcome. Please read DISCLAIMER and COPYING for copyright and license information. Albert Chu chu11@llnl.gov Lawrence Livermore National Laboratory