This document describes the details of the commands and data structures that make up the Containers system. The Containers Guide provides useful context about the workflows and goals of the system that inform these technical details.
This section describes the command line interface to the Containers system.
containerize.py command creates a CREATE experiment made up of containers. The
containerize.py program is available from
users.create.iucc.ac.il. A sample invocation is:
$ /share/containers/containerize.py MyProject MyExperiment ~/mytopology.tcl
It will create a new experiment in MyProject called MyExperiment containing the experiment topology in
mytopology.tcl. All the topology creation commands supported by CREATE are supported by the Containers system, but CREATE program agents are not. CREATE start commands are supported.
Containers will create an experiment in a group if the project parameter is of the form project/group. To start an experiment in the
testing group of the
DETER project, the first parameter is specified as
Names of substrates and nodes in ns2 files are restricted to valid tcl variable names. Names of substrates and nodes in topdl files are restricted to the characters A-Z, a-z, digits, the underscore and the hyphen (-).
containerize.py program will partition the topology into openvz containers, packing 10 containers per physical computer. If the topology is already partitioned - meaning at least one element has a
containers::partition attribute -
containerize.py will not partition it. The
--force-partition flag causes
containerize.py to partition the experiment regardless of the presence of
If container types have been assigned to nodes using the
containerize.py will respect them. Valid container types for the
containers:node_type attribute or the
--default-container parameter are:
containerize.py command takes several parameters that can change its behavior:
Containerize nodes without a container type into kind. If no nodes have been assigned containers, this puts all them into kind containers.
Partition the experiment whether or not it has been partitioned already
Attempt to put int containers into each physical node. The default
Attempt to divide the experiment into int physical nodes. The default is to use packing. There are some nuances to this with mixed containers. See the Containers Guide for more details.
Read configuration variables from filename. Configuration values are discussed below.
Override the site configuration and request nodes of type1 (or type2 etc.) as host nodes.
Attempt to do end node traffic shaping even in containers connected by VDE switches. This works with qemu nodes, but not process nodes. Topologies that include both openvz nodes and qemu nodes that shape traffic should use this. See the discussion below.
Set the default openvz disk space size. The suffixes G and M stand for 'gigabytes' and 'megabytes'.
Set the default openvz template. Templates are described in the Containers Guide.
Add a directory to be searched for openvz templates. Templates must end in
tar.gzand be accessible to the user at creation and swap time. They can only be located under the
Construct a visualization of the virtual topology and leave it in the experiment directories (default).
Ignore network constraints when partitioning nodes.
Do not construct a visualization of the virtual topology.
Specify the packing factor for each partitioning pass. The [ContainersGuide#MoreSophisticatedPacking:MultiplePasses Containers Guide] describes this in detail.
Specify the number of physical machines to use for each partitioning pass. The Containers Guide describes this in detail.
Specify the pnode types on which nodes packed in partitioning pass pass can be placed. The Containers Guide describes this in detail.
Specify the partitioning passes on which network connectivity is ignored. The Containers Guide describes this in detail.
Make sure that Qemu images mount the given users' home directories. Qemu nodes can mount at most 19 users' home directories and this ensures that the experimenters using the containers can reach their home directories.
Print additional diagnostics and leave failed DETER experiments on the testbed.
Do not remove temporary files - used for debugging only.
$ ./containerize.py --packing 25 --default-container=qemu --force-partition DeterTest faber-packem ~/experiment.xml
takes the topology in
~/experiment.xml (which must be a topdl description), packs it into 25 qemu containers per physical node, and creates an experiment called 'DeterTest/faber-packem' that can be swapped-in. If
experiment.xml is already partitioned, it will be re-partitioned. If some nodes in that topology are assigned to openvz nodes already, those nodes will be still be in openvz containers.
The result of a successful
containerize.py run is a CREATE experiment that can be swapped in.
More detailed examples are available in the Containers Guide.
container_image.py command draws a picture of the topology of an experiment. This is helpful in keeping track of how virtual nodes are connected.
containerize.py calls this internally and stores the output in the per-experiment directory (unless
--no-image is used).
A researcher may call
container_image.py directly to generate an image later or to generate one with the partitioning drawn.
The simplest way to call
/share/containers/container_image.py topology.xml output.png
The first parameter is a topdl description, for example the one in the per-experiment directory. The second parameter is the output file for the image. When drawing an experiment that has been containerized, the
--experiment option is very useful.
Draw the experiment in project/experiment, if it exists. Note that this is just the CREATE experiment and project names. Omit any sub-group.
Draw the topology in the filename indicated.
Prefix for containers attributes. Deprecated.
Draw labeled boxes around nodes that share a physical node.
Save the image in the filename indicated.
--program=programname Use programname to lay out the graph. programname must take a file in graphviz's dot language. This is given as the
--programoption to fedd_image.py internally. The default is
fdpwhich works well when
--experiment is given, the first positional parameter is the topdl topology to draw. If
--out is not given the next positional parameter (the first if neither
--experiment is given) is the output file.
A common invocation looks like:
/share/containers/container_image.py --experiment SAFER/testbed-containers ~/drawing.png
Topdl Attributes For Containers
Several topdl attributes influence how an experiment is containerized. These can be added to nodes using the ns2
tb-add-node-attribute command (used throughout the Containers Guide) or directly to the topdl.
These attributes are all attached to nodes/Computers:
The container that will hold this node. The full list is available here.
An identifier grouping nodes together in containers that will share a physical node. Generally assigned by containerize.py, but researchers can also directly assign them. The
containerize.pycommand assigns integers, so if a researcher assigns other partition identifiers,
containerize.pywill not overwrite them.
The flavor of Linux distribution to emulate on openvz. There is a list of valid choices in the Containers Guide.
Amount of disk space to allocate to an openvz container. Be sure to include the G (gigabyte) or M (megabyte) suffix or the size will be taken as disk blocks.
If this attribute is true, resources will be allocated for this node, but it will not be started when the topology is created.
A location to download the QEMU image for this container. The name is a legacy that will disappear. This is deprecated.
There are a few other attributes that are meaningful to more applications. Users specifying ns2 files will not need to set these directly, as the CREATE ns2 interpreter does so.
The start command.
The IPv4 address of this interface. Set by the ns2 commands for fixing addresses.
The IPv4 netmask. ns2 sets this.
These files control the operation of the containers system.
When an experiment is containerized, the data necessary to create it is stored in
/containers. The path
/exp/experiment is created by CREATE when the experiment is created, and used by experimenters for a variety of things. This directory is replicated on nodes under
There are a few files in the per-experiment directory that most experimenters can use:
If the topology was passed to containerize.py as an ns file, this is a copy of that input file. Useful for seeing what the experimenter asked for, or as a basis for new experiments.
The analog of
experiment.tcl, this is the topology given as topdl. The topdl input file.
A drawing of the virtual topology in png format. Generated by container_image.py
The host to IP mapping that will be installed on each node as
site.confA clone of the site configuration file that holds the global variables that the container creation will use. Values overridden on the command line invocation of containerize.py will be present in this file.
The rest of this directory is primarily of interest to developers. It includes:
First version of the input topology after default container types have been added. Input to the partitioning step.
A yaml representation of the partition to virtual node mapping.
The server and channel to use for grandstand communication. Encoded in YAML.
Directory containing the assignment, including all the levels of nested hypervisors.
The contents of the per-experiment directory (except
config.tgz) for distribution into the experiment.
A yaml-encoded representation of the children sub-directory
Containers that are initially not started in the experiment.
Yaml encoding of the qemu images to be used on each node.
Yaml encoding of the openvz templates to be used on each node.
Output of the partitioning process. A copy of
annotated.xmlthat has been decorated with the partitions.
The ns2 file used to create the CREATE experiment.
The topdl file used to generate
The CREATE project and experiment name under which this topology will be created. Broken out into
/var/containers/eidon virtual nodes inside the topology.
A directory containing the routing tables for each node.
Yaml-encoded data about the per-network and per-node loss, delay, and capacity parameters.
A directory containing the VDE switch topology for the experiment.
Yaml-encoded extra switch configuration information. Mostly VDE switch configuration esoterica.
The final topology representation from which the physical topology is extracted. Includes the virtual topology as well. This file can be used as input to container_image.py.
Pickled information for configuring endnode traffic shaping.
Specific parameters for configuring the delay elements in VDE switched topologies that implement traffic shaping. See below.
Site Configuration File
The site configuration file controls how all experiments are containerized across CREATE. The contents are primarily of interest to developers, but researchers may occasionally find the need to specify their own. The
--config parameter to containerize.py does that.
The site configuration file is an attribute-value pair file parsed by a python ConfigParser that sets overall container parameters. Many of these have legacy internal names.
The default site configuration is in
Acceptable values (and their CREATE defaults) are:
Default image used by qemu containers. Default:
Base URL of the CREATE web interface on which users can see experiments. Default:
Hardware used by containers. Default:
Host and port from which to request experiment creation. Default:
OSID to request for qemu container nodess. Default:
Root of the directory tree holding containers software and libraries. Developers often change this. Default:
OSID to request for openvz nodes. Default
Location to load the openvz template from. Default:
True if switched containers (see below) should do traffic shaping in the VDE switch that connects them. Default:
A list of the containers that are networked with VDE switches. Default:
The directory that stores openvz template files. Default:
%(exec_root)s/images/(that is the
imagesdirectory in the
exec_rootdirectory defined in the site config file. This can be a comma-separated list that will be searched in order, after any template directories given on the command line.
The name of the file on experiment nodes used to log containers creation. Default is
The program used to convert ns2 descriptions to topdl. The default is
fedd_ns2topdl.py --filebut any program that takes a single ns2 file as a parameter and prints the topdl to standard output is viable. On CREATE installations
/usr/testbed/lib/ns2ir/parse.tcl -t -x 3 -m dummy dummy dummy dummycan be used to decouple containers from needing a running fedd.
The IP address of a router needed to reach testbed infrastructure
The network on which testbed infrastructure lives that needs to be routed through default_router.
Different container types have some quirks. This section lists limitations of each container, as well as issues in interconnecting them.
Qemu nodes are limited to 7 experimental interfaces. They currently run only Ubuntu 12.04 32 bit operating systems.
These have no way to log in or work as conventional machines. Process tree rooted in the start command is created, so a service will run with its own view of the network. It does not have an address on the control net.
Because of a bug in their internal routing, multi-homed processes do not respond correctly for requests on some interfaces. A ViewOS process does not recognize its other addresses when a packet arrives on a different interface. A picture makes this clearer:
Container A can ping Interface X (10.0.0.1) of the ViewOS container successfully, but if Container A tries to ping Interface Y (10.0.1.2), the ViewOS container will not reply. In fact it will send ARP requests on Interface Y looking for its own address.
For this reason, ViewOS processes are best used as lightweight forwarders.
Physical nodes can be incorporated into experiments, but should only use modern versions of Ubuntu, to allow the Containers system to run their start commands correctly and to initialize their routing tables.
Interconnections: VDE switches and local networking
The various containers are interconnected using either local kernel virtual networking or VDE switches. Kernel networking is lower overhead because it does not require process context switching, but VDE switches are a more general solution.
Network behavior changes such as loss, delay or rate limits are introduced into a network of containers using one of two mechanisms: inserting elements into a VDE switch topology or end node traffic shaping.
Inserting elements into the VDE switch topology allows the system to modify the behavior for all packets passing through it. Generally this means all packets to or from a host, as the Containers system inserts these elements in the path between the node and the switch.
This figure shows three containers sharing a virtual LAN (VLAN) on a VDE switch with no traffic shaping:
The blue containers connect to the switch and the switch has interconnected their VDE ports into the red shared VLAN. To add delays to two of the nodes on that VLAN, the following VDE switch configuration would be used:
The VDE switch connects the containers with shaped traffic to the delay elements, not to the shared VLAN. The delay elements are on the VLAN and delay all traffic passing through them. The Container system configures the delay elements to delay traffic symmetrically - traffic from the LAN and traffic from the container are both delayed. The VDE tools can be configured asymmetrically as well. This is a very flexible way to interconnect containers.
That flexibility incurs a cost in overhead. Each delay element and the VDE switch is a process, do traffic passing from one delayed nodes to the other experiences 7 context switches: container -> switch, switch -> delay, delay -> switch, switch -> delay, delay -> switch, and switch -> container.
The alternative mechanism is to do the traffic shaping inside the nodes, using Linux traffic shaping. In this case, traffic outbound from a container is delayed in the container for the full transit time to the next hop. The next node does the same. End-node shaping all happens in the kernel so it is relatively inexpensive at run time.
Qemu nodes can make use of either end-node shaping or VDE shaping, and use VDE shaping by default. The
--vde-switch-shaping options to
containerize.py force the choice in qemu.
ViewOS processes can only use VDE shaping. Their network stack emulation is not rich enough to include traffic shaping.
Openvz nodes only use end-node traffic shaping. They have no native VDE support so interconnecting openvz containers to VDE switches would include both extra kernel crossings and extra context switches. Because a primary attraction of VDE switches is their efficiency, the Containers system does not implement VDE interconnections to openvz.
Similarly embedded physical nodes use only endnode traffic shaping, as routing outgoing traffic through a virtual switch infrastructure that just connects to its physical interfaces is at best confusing.
Unfortunately, endnode traffic shaping and VDE shaping are incompatible. Because endnode shaping does not impose delays on arriving traffic, it cannot delay traffic from a VDE delayed node correctly.
This is primarily of academic interest, unless a researcher wants to impose traffic shaping between containers using incompatible traffic shaping. There needs to be an unshaped link between the two kinds of traffic shaping.
Bootable Qemu Images
For qemu images to boot reliably, they should not wait for a keypress at the
grub command, which is distressingly common.
To ensure that your image does not wait for
grub, do the following:
For Ubuntu 12.04 (and any system that uses grub2) edit
/etc/default/grub. For example:
GRUB_DEFAULT=0 GRUB_HIDDEN_TIMEOUT=0 GRUB_HIDDEN_TIMEOUT_QUIET=true GRUB_TIMEOUT=1 GRUB_DISTRIBUTOR=`lsb_release -i -s 2> /dev/null || echo Debian` GRUB_CMDLINE_LINUX_DEFAULT="quiet splash" GRUB_CMDLINE_LINUX=""
Just make sure the HIDDENs are not commented out and have true/0 values.
You then must run a command on the system which generates all the new grub configurations:
$ sudo update-grub
The Containers system adds all users to the admin group so that group should be able to use
sudo without providing a password.