PERCEUS Installation, Debian and Debian-based Linux distributions:
[Wrong Distro? RHEL, CentOS, and Scientific Linux quickstart]
The PERCEUS mirror archive is located here: http://mirror.perceus.org/perceus/
Download a source tarball or .deb package. You can use the get-perceus.sh helper script to download, build, and install from source. The instant-debian-vnfs.sh script will create a Debian chroot base environment for you then convert it to a VNFS capsule ready for provisioning.
To build and install from source:
# wget http://mirror.perceus.org/perceus/source/perceus-<version>.tar.gz
# tar xvf perceus-<version>.tar.gz
# cd perceus-<version>
# make install
To install from binary package:
# wget http://mirror.perceus.org/perceus/debs/perceus_<version>_amd64.deb
# dpkg -i perceus_<version>_amd64.deb
Provisioning a cluster with PERCEUS requires that the master server have two network interfaces installed. The primary interface is what the master server uses to install software, system updates, and normal system administration tasks. The secondary interface set with a static IP is used by PERCEUS for a private cluster network that the nodes are provisioned over. The network configuration file for Debian is /etc/network/interfaces, and the secondary interface PERCEUS will use should have settings similar to this:
iface eth1 inet static
When the interfaces config file is saved restart networking or bring up the interface with the “ifup” command.
PERCEUS CONFIG FILES
No changes are necessary here, but check the defaults. In perceus.conf the network interface PERCEUS will use can be set as well as server addresses if different than the PERCEUS master in your setup. The node naming scheme, total node count, or default VNFS capsule can be set in defaults.conf.
INIT & CONFIG SCRIPTS
# perceus init
This is the full system initialization for PERCEUS, and it will be forced to run if another PERCEUS command is tried before the system is initialized. Default settings will work for provisioning a cluster. (Check the values before accepting just in case.) It will give an error if the PERCEUS network interface has not been configured.
# perceus configure hosts
This will populate the master's /etc/hosts file with entries for the cluster nodes. These will be used for static IP addressing when provisioning nodes, and will also be used by the hostname, ipaddr, and hostfile PERCEUS modules.
PERCEUS MODULE ACTIVATION
# perceus module states
This will show the imported PERCEUS modules and their activated states. A common recommended group of modules for many clusters (and their activated states) is:
(NOTE: No modules are necessary to provision a cluster but they do add helpful functionality, and can be customized or created.)
IMPORT VNFS CAPSULE
# perceus vnfs import <capsule name>.vnfs
You will need to have downloaded a VNFS capsule from the PERCEUS mirror (http://mirror.perceus.org/perceus/vnfs/) or built one using the VNFS helper scripts. After importing, you can add the VNFS capsule name to the /etc/perceus/defaults.conf file or assign it to existing cluster nodes with the command:
# perceus node set vnfs <capsule name>
To see the names of imported VNFS capsules:
# perceus vnfs list
To see which VNFS capsules are assigned to which cluster nodes:
# perceus node summary
These may start by default, if not run these commands:
# /etc/init.d/nfs-kernel-server start
# /etc/init.d/perceus start
If there is an error about ports in use, check the notes section for troubleshooting help.
Power on nodes and watch for them to PXE boot. Stage 1 provisioning has a boot splash menu with options to provision, boot from internal hard disk, or provision with debug messages. Past this menu (into stage 2 of provisioning) if presented with an error like "no compatible network interfaces found" you will need to use an updated TFTP boot image image that has driver support for your hardware. Also, contact us and let us know what your unsupported network interface is so we can add support for it in the PERCEUS kernel.
If it is reported that no VNFS capsule is set for the node, run:
# perceus node set vnfs <capsule name> <node name>
This will assign a VNFS to the node(s) and it should immediately continue.
If the node boots to a login prompt provisioning is successful. If the boot messages for nodes throw errors there may be a problem with the VNFS capsule itself; try downloading and provisioning a premade VNFS, use a different VNFS build script or modify the one that was used. Again, if encountering errors please let us know so they can be fixed.