wiki:NeoContainers

Version 45 (modified by Geoff Lawler, 8 years ago) (diff)

--

Neo-Containers -- DETER Containers

Section List

  1. The Basics
  2. Initial Set-up
  3. Using Neo-Containers with the Existing Containers System
  4. Using Neo-Containers While Bypassing the Existing Containers System
  5. After Booting the Containers
  6. Experiment Completion
  7. Extensions to Neo-Containers

1. The Basics

The Neo-Containers system uses cloud-container technology to abstract and generalize container creation and initialization. At the DETER level, the experiments have a number of physical nodes, "pnodes", which serve as hosts for the virtualized containers. Outside of the experiment there are two servers which configure the containers. Both run on chef.isi.deterlab.net. The Chef server serves "code as configuration" and stores static (for the most part) configuration information. The config_server is a RESTful API which loads then serves experiment-specific configuration information. The code that the Chef server runs on the containers usually pull the specific configuration information from the config_server.

The config_server code can be found on DETER's Github account in the config_server repository. The Chef recipes used are there as well in the deter-chef repository. (Although the deter-chef repository is currently private to DETER dev and OPs.)

There are two approaches to running configuring virtual machines in Neo-Containers: leveraging the existing DETER containers system or not. These can be mixed and matched; both can be used to describe the final node and network configuration of the containers in the experiment. If you use the existing containers system. please refer to the DETER Containers Docuementation for details it. If you want to add "non-containerized" nodes to your experiment, you'll have to write a simple JSON formatted configuration file which describes the virtual machines you want to add (IP addresses, hostnames, OS, etc).

2. Adding virtual machines "by hand" to your experiment.

Create a nodes.json File

A nodes.json file must be created that will describe the containers added to the experiment. This file is only used to define the containers for this experiment. (The file need not be named nodes.json, but that is the name that will be used in this documentation.)

The file is a list of node configurations, where each node configuration must have the following fields defined:

  • name - This is the name of this container.

  • host - This is the pnode in the experiment on which this container will run. Needless to say, this node must exist in the experiment.

  • interfaces - This a list of the network interfaces to be assigned the container.

  • image_os - This is really just to distinguish Windows from non-Windows nodes. If the image_os starts with "windows", the image will be treated as a Windows node. Otherwise it'll be assumed to be some sort of Unix-y container.

  • image_type - This setting describes the containerization tech of the node. Currently this is *always* set to "vagrant" as Vagrant is the only package used to spin up the containers.

  • image_name - The name of the image. Any containers that share a name will also share an image.

  • image_url - A URL must be specified which the neo-containers system uses to download the container image. This URL must be resolvable from the experiment nodes. The image will only be downloaded once as long as the image_names are the same for each container. Existing and supported images are Ubuntu 14.04 64 (at http://scratch/containers/deter_ub1404_64_vb.box) and Windows 7 (at http://scratch/containers/deter_win7.box).

The following is an example nodes.json file that creates one Ubuntu 14.04 container and one Windows container:


    [
        {
            "host": "leda",
            "name": "sarah",
            "image_url": "http://scratch/containers/deter_ub1404_64_vb.box",
            "image_os": "ubuntu 14.04 64",
            "image_type": "vagrant",
            "image_name": "deter/ub14",
            "interfaces": [
                { "address": "10.1.1.101", "mac": "de:ad:be:ef:00:be" }
            ]
        },
        {
            "host": "swan",
            "name": "helena",
            "image_url": "http://scratch/containers/deter_win7.box"
            "image_os": "windows",
            "image_type": "vagrant",
            "image_name": "deter/win7",
            "interfaces": [
                { "address": "10.1.1.201", "mac": "de:ad:be:ef:00:af" }
            ]
        }
    ]
Note that this is a list of configurations. In JSON, a list is a comma-separated list of times between square brackets like so: [ item, item ..., item ]. Each configuration block is given as a dictionary with the key/value pairs as above. A JSON dictionary is delimited by { }. So a list of dictionaries is [ { ... }, { ... }, ... ].

3. Tell the Configuration Server about your new Nodes

The Configuration Server needs to know that information in your brand-new nodes.jsonfile. There's a little script just for this: /share/config_server/bin/initialize_containers.py. This script feeds the information from your nodes.json file to the Configuration Server. This must be done before you swap in your experiment for reasons given below. (Note that because the script needs to run before the experiment is swapped in, the script is run on the users machine.

The script asks DETER to allocate control network addresses for your new nodes. These addresses must exist prior to the containers existing as the control network to request configuration information from the configuration server. And due to the way DETER works, the addresses must be allocated before swap in or the addresses will not be properly associated with the containers hostnames. The hostnames are how the Configuration Server talks to your containers. So if the script is not run, the configuration of the new nodes cannot happen. The script can be run multiple times without ill effects. The system is smart enough to only request the control network addresses from DETER once. If the script is re-run, the (non-control-net-IP) information will simply overwrite the existing information in the configuration database.

Below is an example of how to run the script. The <expid> and <projid> fields in the example refer to the experiment ID and the project ID. The experiment ID is defined by the user, and could be something like "neocont-test" or "netstriping". The project ID is the name of the project under which the experiment is run.

    $ /share/config_server/bin/initialize_containers.py -p <projid> -e <expid> -f path/to/nodes.json
When run, you'll see output in the terminal. If successful, you will not see any [ERROR]s. [WARN ]s may be OK. Here's sample output: [PLACEHOLDER FOR IMAGE]

3. Using Neo-Containers with the Existing Containers System

This method of using Neo-Containers uses the existing Containers system. This method allows the use of more complex network topologies.

Create an Experiment

Create an experiment using the existing Containers system. An NS file and the /share/containers/containerize.py script are used to create the containerized experiment.

In your NS file for each container, specify image_os, image_type, image_name, and image_url via the tb-add-node-attribute syntax. Details on each attribute are given below.

  • image_os - This is really just to distinguish Windows from non-Windows nodes. If the image_os starts with "windows", the image will be treated as a Windows node. Otherwise it'll be assumed to be some sort of Unix-y container.

  • image_type - This setting describes the containerization tech of the node. Currently this is *always* set to "vagrant" as Vagrant is the only package used to spin up the containers.

  • image_name - The name of the image. Any containers that share a name will also share an image.

  • image_url - A URL must be specified which the neo-containers system uses to download the container image. This URL must be resolvable from the experiment nodes. The image will only be downloaded once as long as the image_names are the same for each container. Existing and supported images are Ubuntu 14.04 64 (at http://scratch/containers/deter_ub1404_64_vb.box) and Windows 7 (at http://scratch/containers/deter_win7.box).

The following is an example NS file that creates one Windows container and one Ubuntu 14.04 container:

    set r2d2 [$ns node]
    tb-add-node-attribute $r2d2 containers:image_os windows
    tb-add-node-attribute $r2d2 containers:image_type vagrant
    tb-add-node-attribute $r2d2 containers:image_name deter/win7
    tb-add-node-attribute $r2d2 containers:image_url
    http://scratch/containers/deter_win7.box

    set c3po [$ns node]
    tb-add-node-attribute $c3po containers:image_os ubuntu
    tb-add-node-attribute $c3po containers:image_type vagrant
    tb-add-node-attribute $c3po containers:image_name ubuntu/trusty64
    tb-add-node-attribute $c3p0 containers:image_url
    http://scratch/containers/deter_ub1404_64_vb.box

Containerize the Experiment

Use the NS file to create a containerized experiment using the existing Containers scripts.

    $ /share/containers/containerize.py 

Note: The experiment must currently be created in the Deter group as that's where the custom pnode disk images are. This will change.

Finalize the NS File

Modify the NS file generated by containerize.py to have a new image for the pnode machines.

Follow these steps in your browser:

  1. Go to the new experiment page.
  2. Click Modify Experiment.
  3. Remove all existing tb-set-node-startcmd lines.
    These start the old Containers system and are no longer used.
  4. For each pnode, change the OS type to PNODE_BASE.
  5. For each pnode, change the hardware type to MicroCloud.

After making these modifications, each pnode in the NS file should have these lines:

    tb-set-node-os ${pnode(0000)} PNODE-BASE
    tb-set-hardware ${pnode(0000)} MicroCloud

The final NS file will look something like this:

    set ns [new Simulator]
    source tb_compat.tcl

    tb-make-soft-vtype container0 {dl380g3 pc2133 MicroCloud}
    set pnode(0000) [$ns node]
    tb-set-node-os ${pnode(0000)} PNODE-BASE
    tb-set-hardware ${pnode(0000)} container0
    tb-set-node-failure-action ${pnode(0000)} "nonfatal"

    $ns rtproto Static
    $ns run

Swap In

On the experiment's webpage, swap in the experiment.

Populate the Configuration Database

Populate the configuration database that runs on chef.isi.deterlab.net by running the load_containers_db.sh and load_config_db.sh database-population scripts.

This should be run on a single physical node in the experiment. pnode-0000 is used in the example below.

The <expid> and <projid> fields in the following example are referring to the experiment ID and the project ID. The experiment ID is defined by the user, and could be something like "neocont-test" or "netstriping". For now, the project ID should always be "Deter".

    $ ssh pnode-0000.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ ./load_config_db.sh
    $ ./load_containers_db.sh -p <projid> -e <expid>

This step will be automated in the future.

Node Configuration by Chef

The Chef system is used to bootstrap and configure the nodes. All the steps for this are enclosed in the bootstrap_node.sh script.

The script needs to know which node's role in the experiment. There are currently three roles: pnode, container, and win-container.

On all the pnodes which will be running containers:

    $ ssh <pnode>.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ ./bootstrap_node.sh -r pnode

The pnode only have to be bootstrapped once per experiment swap in. Once a pnode is bootstrapped into chef, chef-client needs to be run. The pnode role will spawn the containers and configure them. So once the chef-client command is run on a pnode, all containers on that be pnode will be running and configured.

    $ ssh <pnode>.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ sudo chef-client

It is easy to fix problems if something should go wrong with bootstrapped nodes. Running "sudo chef-client" will re-configure the nodes (both pnodes and the containers).

Set-up Complete

If all the preceding steps succeeded, then your pnodes and containers are configured, booted, and ready for use.


Swap In

On the experiment's webpage, swap in the experiment.

Populate the Configuration Database

Populate the configuration database that runs on chef.isi.deterlab.net by running the load_containers_db.sh and load_config_db.sh database-population scripts.

This should be run on a single physical node in the experiment. pnode-0000 is used in the example below.

The <expid> and <projid> fields in the following example are referring to the experiment ID and the project ID. The experiment ID is defined by the user, and could be something like "neocont-test" or "netstriping". For now, the project ID should always be "Deter".

    $ ssh pnode-0000.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ ./load_config_db.sh

This step will be automated in the future.

Node Configuration by Chef

The Chef system is used to bootstrap and configure the nodes. All the steps for this are enclosed in the bootstrap_node.sh script.

The script needs to know which node's role in the experiment. There are currently three roles: pnode, container, and win-container.

On all the pnodes which have containers running on them:

    $ ssh <pnode>.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ ./bootstrap_node.sh -r pnode
The pnode only have to be bootstrapped once per experiment swap in. Once a pnode is bootstrapped into chef, chef-client needs to be run. The pnode role will spawn the containers and configure them. So once the chef-client command is run on a pnode, all containers on that be pnode will be running and configured.
    $ ssh <pnode>.<expid>.<projid>
    $ cd <config_server-repo>/bin
    $ sudo chef-client

It is easy to fix problems if something should go wrong with bootstrapped nodes. Running "sudo chef-client" will re-configure the nodes (both pnodes and the containers).

Set-up Complete

If all the preceding steps succeeded, then your pnodes and containers are configured, booted, and ready for use.

5. After Booting the Containers

There are a number of things that may be done after the containers are configured and booted. These include the following:

  • Login to a container node
        $ ssh pnode
        $ ssh username@containernode
    
    Cygwin is installed on Windows nodes so you can ssh to Windows containers as well.

  • Play around with Vagrant
    Vagrant is used to control and access the container nodes. Vagrant is used from a root shell ("sudo su -") and from the /space/vagrant_home directory.

    Command Purpose
    vagrant status confirm containers are running
    vagrant ssh containernode login "vagrant"; password "vagrant"
    vagrant reload containernode reboot a container
    vagrant halt containernode halt a container
    vagrant up containernode boot a container

  • Login to a Windows desktop
    • build ssh tunnel to port 3389 on the pnode
      $ ssh -L3389:pcXXX:3389 users.deterlab.net

    • use client RDP to connect to localhost:3389
      login "vagrant"; password "vagrant" for an Administrator shell or login [your user id]; password "password" for normal user access.

6. Experiment Completion

After an experiment is complete, the experiment data must be removed from the configuration database. There are two ways this may be done.

Method 1: On a host which can talk to chef.isi.deterlab.net, run these commands:

    $ cd <config_server-repo>/bin
    $ rm_experiment_config.sh -p <projid> -e <expid>

Method 2: The config_server may be called directly:

    $ curl http://chef:5320/exp/<projid>/<expid>/delete

7. Extensions to Neo-Containers

Chef Workstation

Since the system runs on Chef anyone authorized to push chef recipes to the chef server can write custom node configuration code.

Vagrant

The system uses Vagrant to spin up the containers, thus any Vagrant-supported image can run in a container. However, the image must be downloaded and served via HTTP.


Windows 7 Example

This example shows how to bring up Win7 nodes. It does not use the old containers system. It has two nodes in a single LAN, then spins up a single Win7 node on each.

NS file with two nodes, both running the PNODE-BASE image.

set ns [new Simulator]
source tb_compat.tcl

tb-make-soft-vtype container0 {pc2133 MicroCloud dl380g3}

set base1 [$ns node]
tb-set-node-os $base1 PNODE-BASE
tb-set-hardware $base1 container0
tb-set-node-failure-action $base1 "nonfatal"

set base2 [$ns node]
tb-set-node-os $base2 PNODE-BASE
tb-set-hardware $base2 container0
tb-set-node-failure-action $base2 "nonfatal"

set lan1 [$ns make-lan "$base1 $base2" 100Mb 0.0ms DropTail]
tb-set-ip-lan $base1 $lan1 10.0.0.1
tb-set-ip-lan $base2 $lan1 10.0.0.2

$ns rtproto Static
$ns run

Create a new experiment with the NS file above. *Do not* swap in the experiment.

We run our own configuration database on the chef node. This is not required, but keeps our configuration separate from the "official" configuration database.

users: > ssh chef
chef: > cd src/config_server
chef: > git pull ~chef-user/config_server   # grab the newest code from chef-user.
chef: > ./runserver.py -l debug -p 5323 -d $(pwd)/config.db    # full path to DB and unused port.

Create a nodes.json file which describes the Win7 nodes to create:

[
        {
                "host": "base1",
                "name": "billg",
                "image_url": "http://scratch/containers/deter_win7.box",
                "image_os": "windows",
                "image_type": "vagrant",
                "image_name": "deter/win7",
                "interfaces": [
                    { "address": "10.0.0.101", "mac": "de:ad:be:ef:00:bf" }
                ]
        },
        {
                "host": "base2",
                "name": "ballmer",
                "image_url": "http://scratch/containers/deter_win7.box",
                "image_os": "windows",
                "image_type": "vagrant",
                "image_name": "deter/win7",
                "interfaces": [
                    { "address": "10.0.0.102", "mac": "de:ad:be:ef:00:c1" }
                ]
        }
]

Note that the Win7 nodes are in the same subnet at the physical nodes.

Load the configuration into our configuration server running on chef. This must be done *before* swapping in the experiment as the configuration server must request control network addresses from DETER.

users: > /share/config_server/bin/initialize_containers.py -p deter -e neo-win7 -f neo-win7.nodes.json -P 5323 -l debug

Note that neo-win7.nodes.json is the JSON file above and we use the port of our own config_server.

Output should not contain [ERROR] or [WARN] messages. Here's an example of a good run:

01-27 09:29:11 __main__     INFO     Log level set to debug
01-27 09:29:11 __main__     DEBUG    using base url: /exp/deter/neo-win7/insert
[OK   ] : Initializing containers from neo-win7.nodes.json for use in experiment deter/neo-win7
01-27 09:29:11 __main__     DEBUG    sending url: http://chef:5323/exp/deter/neo-win7/insert/containers
[OK   ] : upload_node_file: Connecting to server chef:5323
[OK   ] : upload_node_file: Posting node data ... url=/exp/deter/neo-win7/insert/containers
[OK   ] : Upload_node_file: Getting Response
[OK   ] : Good response from server. Message: Nodes added: billg, ballmer
[OK   ] : All is well. Nodes initialized.

Now swap in the experiment:

users: > swapexp -wNe deter,neo-win7 in

Once swapped in, we log into a base node and kick off the chef processes to create and start the Win7 node.