eucalyptus.conf, eucalyptus.local.conf - configuration files for
This manual describes the configuration files used to configure and
/etc/eucalyptus/eucalyptus.conf is the traditional Eucalyptus
configuration file. In previous Eucalyptus versions (<1.6.2), and on
non-Ubuntu versions, Eucalyptus is configured by this one monolithic
file, with inline comments describing all options. As of Ubuntu's
Eucalyptus 1.6.2 version, the inline documentation has been extracted
to two manpages, euca_conf(8) and this page, eucalyptus.conf(5).
/etc/eucalyptus/eucalyptus.conf now contains a base set of your
distribution's recommended configuration. This file is managed by
dpkg(1), and provides a vector for your distribution to add new
variables/values and maintain a sane set of defaults. It serves as the
base configuration, which are easily overriden by the administrator.
/etc/eucalyptus/eucalyptus.local.conf contains values written by
euca_conf(8). This file should never be edited directly. Rather, the
user should see the documentation at euca_conf(8) and use the various
options and interfaces for modifying this file.
Each option below may affect one or more of the following Eucalyptus
- CLC = Cloud Controller
- WC = Walrus Controller
- CC = Cluster Controller
- SC = Storage Controller
- NC = Node Controller
**NOTE**: To activate changes of any parameters on a CC, you must:
sudo restart eucalyptus-cc CLEAN=1
**HOWEVER**, if you do this, all currently running virtual machines in
this cluster will lose network connectivity.
Options affecting CLC, WC, CC, SC, NC:
This variable points to where eucalyptus has been installed.
This is the username that you would like eucalyptus to run as.
Options affecting CLC, WC, SC only:
Enable/disable Eucalyptus dynamic DNS functionality.
Use either iSCSI or AoE for dynamic block storage.
Amount of memory to allocate to the JVM. This number value will
affect how many instances you can over a period of time. These
processes are memory hungry; give them as much memory as you can
spare. At least 2048m is recommended for very busy Eucalyptus
Options affecting CC, NC only:
This variable controls whether ws-security is enabled between
eucalyptus components. The default settings provide secure
connections between the Cloud, Cluster, and Node Controllers and
we recommend that this feature remains enabled. If you wish to
disable security, you must change this variable to "N" and
manually configure the services.xml for both Cluster and Node
Controllers (see documentation for more details).
This variable controls the level of logging output that appears
in various eucalyptus log files. The options are, in descending
order of verbosity, 'DEBUG, INFO, WARN, ERROR, and FATAL'. The
default is DEBUG (everything).
The 2 variable VNET_PRIVINTERFACE and VNET_PUBINTERFACE specify
the local physical ethernet interfaces that eucalyptus should
use to manage the VM network. On the front-end,
VNET_PRIVINTERFACE should be set to the device that is attached
to the same ethernet network as your nodes. - VNET_PUBINTERFACE
should be set to the device which is connected to the 'public'
network. If you have only one interface, these should be set to
the same value. On the nodes, both should be set to either the
name of the bridge that has been set up by Xen (xenbr0, eth0,
etc), or the physical ethernet device that is attached to the
xen bridge (peth0, peth1, etc), depending on your xen
configuration. CAVEAT: These parameters in eucalyptus.conf do
not affect Storage Controller interface settings--to do this,
please use the administrative web interface.
There are four modes to choose from (MANAGED, MANAGED-NOVLAN,
SYSTEM, or STATIC) and each has its own sub-options. The first
modes (MANAGED, MANAGED-NOVLAN) configure eucalyptus to fully
manage the VM networks, and enables the ability to use security
groups and dynamic public IP assignment (with and without vlan
tagging of security group networks, respectively).
Options affecting CC only:
This is the port the Cluster Controller will be listening on.
This option configures the Cluster Controller's scheduling
policy. Currently, this option can be set to GREEDY (first node
that is found that can run the VM will be chosen), ROUNDROBIN
(nodes are selected one after another until one is found that
can run the VM), or POWERSAVE (nodes are put to sleep when they
are not running VMs, and reawakened when new resources are
required. VMs will be placed on the first awake machine,
followed by machines that are asleep).
Powersave options. POWER_IDLETHRESH is the number of seconds
that a node can remain idle (i.e. no running VMs) before a
powerdown is attempted. POWER_WAKETHRESH is the number of
seconds that Eucalyptus should wait after attempting a node
wake-up before it will consider the node actually down (and not
The name of the Node Controller service. Change this if you want
to plug in your own Node Controller service.
This indicates where we have a dhcp server binary. We use it to
provide the images with IPs: Eucalyptus provides its own
configuration per instance.
Some systems have their DHCP daemon configured to run as a non-
root user. If this is the case, set the name of that user here
(by default, Eucalyptus will set up DHCPD configuration files
and directories as owned by root).
The list of Node Controllers the Cluster Controller will
communicate with. If you are running Rocks, you can run "rocks
list host" to find out the list of machines available to you (in
our case we are interested in the VM Container kind).
VNET_SUBNET should be set to an IP subnet that is free for
eucalyptus to use (i.e. no other system connected to your
network directly is configured with addresses from this subnet).
VNET_NETMASK defines the size of the subnet.
VNET_DNS should be set to a DNS server that your systems use
(usually safe to use the same DNS that is configured on the
VNET_ADDRSPERNET can be used to limit the number of instances
that can be attached to each named security group
VNET_PUBLICIPS should be set to any public IPs, that are
currently unused, that can be dynamically assigned to VMs. Of
these options, only VNET_PUBLICIPS can be left blank or
If you are running in multi-cluster mode (more than one CC), you
should uncomment VNET_LOCALIP and set it to the local IP of the
CC that is accessible by all other CCs in the system. If
VNET_LOCALIP is unset, the CC will try to determine the list of
all IPs currently assigned to the machine at CC run time.
If your CC and CLC are on different machines, uncomment
VNET_CLOUDIP and set it to your cloud-contoller's IP address
(must be an address that can be reached by the CC).
Options affecting the NC only:
This is the port the Node Controller will be listening on.
The hypervisor that the Node Controller will interact with in
order to manage virtual machines. Currently, supported values
are 'kvm' and 'xen'.
Setting this to 1 disables the cleanup of instance files (root,
kernel, ramdisk) for failed and terminated instances. This is
not recommended for normal use, but it can be useful in
debugging VM startup.
The maximum amount of disk space, in Megabytes, that Eucalyptus
is allowed to use in the cache directory
(INSTANCES_PATH/eucalyptus/cache). A generous size is
recommended. Setting this to zero disables caching.
VNET_BRIDGE should be set to the name of the bridge that xen has
configured. This is typically named 'xenbr0, xenbr1, etc' on
older Xen versions, and 'eth0, eth1, etc' on newer Xen versions.
The command 'brctl show' will give you more information on your
local bridge setup.
This variable points to a directory which is used by the Node
Controller to store images of running instances as well as local
cached copies of images. The running images will be deleted
after the instance is terminated, but the cached copies will
persist, subject to LRU cache replacement and the NC_CACHE_SIZE
size limit, below. So, this partition should be at least as big
as the cache size (or the maximum space needed by all images,
whichever is bigger) plus the maximum space needed by the
maximum number of instances allowed on the node. This directory
should be local to the Node Controller (as opposed to a NFS
share) for performance reasons.
The maximum amount of memory Eucalyptus is allowed to use on the
node: if you leave this commented out, Eucalyptus will use all
available memory, otherwise it will use at most this value for
ALL running instances.
The maximum number of CPU/cores Eucalyptus is allowed to use on
the node (at the moment we don't differentiate between cores and
CPU). If you leave this commented out, Eucalyptus will use all
available CPU/cores it can find.
The size of the swap partition, in MB, for each instance started
on the node (default is 512MB). If the maximum disk allowed for
the instance is not big enough to accommodate the swap together
with the root partition, then no swap is allocated. If there is
extra room left, then an "ephemeral" partition will be created,
available as /dev/sda3 inside the VM.
Set the number of 'disk intensive operations' that are allowed
to take place in parallel on this machine. If this value is set
too high, we run the risk of driving the load on the machine too
high when multiple instances are started on this single node.
The default is 4.
This manpage was written by Dustin Kirkland <firstname.lastname@example.org>
(with extensive help from Daniel Nurmi, and some of it lifted from the
original /etc/eucalyptus/eucalyptus.conf) for Ubuntu systems (but may
be used by others). Permission is granted to copy, distribute and/or
modify this document under the terms of the GNU General Public License,
Version 3 published by the Free Software Foundation.
On Debian systems, the complete text of the GNU General Public License
can be found in /usr/share/common-licenses/GPL.