teleport/docker
Sasha Klizhentas ebe3c1a9c1 Fix several issues with audit events.
1. Fixes several cosmetic issues with logs:

Fixes #1690, fixes #1687

2. Fixes deadlocks that were revealed during stress
testing on slow encrypted EFS system.

The following deadlock scenario was happening:

Goroutine 1:

t1. auditlock.Lock <- success
t3. diskSessionLogger.Lock <- blocked

Gorotuine 2:

t2. diskSessionLogger.Lock <- success
t4. auditLock.Lock  <- blocked

3. Update ansible upgrade scripts
to support custom binary upgrades.

4. Fix docker flow by removing deprecated
--gops-addr flag

5. Remove verbose logging lines.

6. Reduce compression efficiency to
reduce large memory load.
2018-02-15 18:28:42 -08:00
..
ansible Fixed recursive copy path. 2017-04-19 12:02:17 -07:00
.bashrc Updated docker-based example 2017-02-23 21:45:13 -08:00
.dockerignore Created Docker-based test bed 2017-02-18 22:06:12 -08:00
.env refactoring 2017-10-05 17:29:31 -07:00
.gitignore Created Docker-based test bed 2017-02-18 22:06:12 -08:00
.screenrc Minor dockerile/make changes 2017-02-23 17:00:56 -08:00
docker-compose.yml Fix several issues with audit events. 2018-02-15 18:28:42 -08:00
Dockerfile Mutual TLS Auth server and clients. 2017-12-27 11:37:19 -08:00
env.file refactoring 2017-10-05 17:29:31 -07:00
Makefile Allow to pick pro or enterprise mode in docker flow 2018-01-09 17:56:08 -08:00
one-node.yaml fix delays and offsets, address review comments 2017-11-16 14:43:35 -08:00
one-proxy.yaml refactoring 2017-10-05 17:29:31 -07:00
one.yaml Add support for NFS-friendly log protocol. 2018-01-04 18:54:37 -08:00
README.md Updated trusted cluster documentation for docker build. 2017-04-11 16:57:59 -07:00
two-auth.yaml refactoring 2017-10-05 17:29:31 -07:00
two-node.yaml refactoring 2017-10-05 17:29:31 -07:00
two-proxy.yaml refactoring 2017-10-05 17:29:31 -07:00
two-role-admin.yaml Fixed Docker admin role. 2017-04-19 12:02:17 -07:00
two-tc.yaml more work, discovery works 2017-10-07 18:11:03 -07:00

Docker

This directory contains Docker-based flow to run Teleport clusters locally for testing & development purposes.

Building

First, you need to build teleport:latest Docker image. This image is built automatically when you type make BUT...

But you have to build the base image first, by running make -C build.assets from $GOPATH/github.com/gravitational/teleport (repository base dir).

Starting

Type:

$ make

This will start two Teleport clusters:

Stopping

Type:

$ make stop

Configuration

Look at the Makefile: the containers are started with their /var/lib/teleport mounted to data/one or data/two on a host.

The configuration is passed via YAML files located in /teleport/docker/xxx.yaml inside each container.

Since the cluster data is preserved between restarts, so you can edit the configuration and restart if you want to change configuration changes.

Using TCTL

To add users to any of the clusters, you have to "enter" into the running containers of their auth servers and use tctl there.

For cluster "one":

$ make enter-one

and then you'll find yourself inside a container where teleport auth daemon is running, try ps -ef for example and you'll get something like this:

container(one) /teleport: ps -ef
UID        PID  PPID  C STIME TTY          TIME CMD
root         1     0 40 06:04 ?        00:00:06 build/teleport start -c /teleport/docker/one.yaml
root        13     0  0 06:04 ?        00:00:00 /bin/bash
root        19    13  0 06:04 ?        00:00:00 ps -ef

For cluster "two":

$ make enter-two

... and then you can use stuff like tctl users add, etc. Make sure to pass the YAML file to tctl via -c flag.

Trusted Clusters

Trusted Clusters with Resources

  1. Update two-role.yaml and replace username_goes_here with your username.

  2. Create a Role and TrustedCluster resource on Cluster Two.

    make enter-two
    tctl -c /root/go/src/github.com/gravitational/teleport/docker/two-auth.yaml create -f docker/two-role-admin.yaml
    tctl -c /root/go/src/github.com/gravitational/teleport/docker/two-auth.yaml create -f docker/two-tc.yaml
    

Trusted Clusters with File Configuration

Export CAs

Run the following commands to export your CAs.

# enter cluster two and export ca
make enter-two
tctl -c /root/go/src/github.com/gravitational/teleport/docker/two-auth.yaml auth export > docker/data/two/two.ca
exit

# enter cluster one and export ca
make enter-one
tctl auth export > docker/data/one/one.ca
exit
Upate Configuration

Stop both clusters with make stop, update the file configuration for both clusters, and start again with make.

# update docker/one.yaml with the following under "auth_service"
trusted_clusters:
  - key_file: /root/go/src/github.com/gravitational/teleport/docker/data/two/two.ca
# update docker/two-auth.yaml with the following under "auth_service"
trusted_clusters:
  - key_file: /root/go/src/github.com/gravitational/teleport/docker/data/one/one.ca
    allow_logins: root
    tunnel_addr: one

Ansible

To setup Ansible:

  1. Follow steps in Trusted Cluster section to setup Trusted Clusters.

  2. Use tctl to issue create user command and follow link on screen to create user.

    tctl users add {username} root
    
  3. Configure Ansible.

    # add two-node to ansible hosts file
    echo "172.10.1.2:3022" >> /etc/ansible/hosts
    
    # setup ssh_args that ansible will use to access trusted cluster nodes
    sed -i '/ssh_args = -o ControlMaster=auto -o ControlPersist=60s/assh_args = -o "ProxyCommand ssh -p 3023 one -s proxy:%h:%p@two"' /etc/ansible/ansible.cfg
    
    # use scp over sftp
    sed -i '/scp_if_ssh/s/^#//g' /etc/ansible/ansible.cfg
    
  4. Start and load OpenSSH agent with keys.

    # create directory for ssh config
    mkdir ~/.ssh && chmod 700 ~/.ssh
    
    # start ssh-agent
    eval `ssh-agent`
    
    # log in with the user created before
    tsh --proxy=localhost --user=rjones login
    
    # load keys into ssh-agent
    tsh --proxy=localhost --user=rjones agent --load
    
  5. Verify Ansible works:

    $ ansible all -m ping
    172.10.1.2 | success >> {
        "changed": false, 
        "ping": "pong"
    }
    
  6. Run an simple playbook:

    # cd to directory that contains playbook
    cd /root/go/src/github.com/gravitational/teleport/docker/ansible
    
    # run playbook
    ansible-playbook playbook.yaml
    

Interactive Usage

Also you can start an empty container from which you can manually invoke teleport start. This is similar to launching an empty Linux VM with a Teleport binary.

To get shell inside the same "one" (single-node cluster) container without Teleport running:

$ make shell

NOTE: If you get "network already exists" error, do make stop first.

Once inside, you'll get the same /var/lib/teleport as "one", so you can start (and even build) teleport daemon manually. This container also comes with a fully configured screen so you can treat it as a real VM.