4627a92447
#15563 - [X] Manual QA for all new/changed functionality Tested by running the following: If the changes haven't been merged to `main`: ```sh fleetctl preview --preview-config 15563-move-external-dep-osquery-in-a-box-to-monorepo fleetctl preview stop fleetctl preview reset ``` If the changes were already merged to `main`: ```sh fleetctl preview fleetctl preview stop fleetctl preview reset ``` |
||
---|---|---|
.. | ||
in-a-box | ||
.env | ||
docker-compose.linux-overrides.yml | ||
docker-compose.yml | ||
example_osquery.flags | ||
fleet.crt | ||
fleet.key | ||
README.md |
Configs and Tools for Testing Fleet
The files in this directory are intended to assist with Fleet development.
-
docker-compose.yml
: This docker-compose file helps with startingosqueryd
instances for testing Fleet. More on this below. -
example_osquery.conf
: An example osquery config file that sets up basic configuration for distributed queries. -
example_osquery.flags
: An example osquery flagfile setting the config options that must be loaded before the full JSON config. -
fleet.crt
&fleet.key
: Self-signed SSL certificate & key useful for testing locally withosqueryd
. Works with the domainhost.docker.internal
(exposed within docker containers as the host's IP). Should never be used in production.
Testing with containerized osqueryd
Using Docker enables us to rapidly spin up and down pre-configured osqueryd
instances for testing Fleet. Currently we have container images for Ubuntu14 and Centos7 osquery installations.
Setup
Docker and docker-compose are the only dependencies. The necessary container images will be pulled from Docker Cloud on first run.
Set the environment variable ENROLL_SECRET
to the value of your Fleet enroll secret (available on the manage hosts page, or via fleetctl get enroll-secret
).
(Optionally) Set FLEET_SERVER
if you want to connect to a fleet server
besides host.docker.internal:8080
.
Running osqueryd
The osqueryd instances are configured to use the TLS plugins at host.docker.internal:8080
. Using the example_osquery.flags
in this directory should configure Fleet with the appropriate settings for these osqueryd
containers to connect.
To start one instance each of Centos 6, Centos 7, Ubuntu 14, and Ubuntu 16
osqueryd
, use:
docker-compose up
Linux users should use the overrides (which add DNS entries for
host.docker.internal
based on the DOCKER_HOST
env var):
docker-compose -f docker-compose.yml -f docker-compose.linux-overrides.yml up
The logs will be displayed on the host shell. Note that docker-compose up
will reuse containers (so the state of osqueryd
will be maintained across calls). To remove the containers and start from a fresh state on the next call to up
, use:
docker-compose rm
If you want to only start one instance of osqueryd
, use:
docker-compose run ubuntu14-osquery
or
docker-compose run centos7-osquery
Note that docker-compose run
does not save state between calls.
This system can also be used to start many instances of osqueryd running in containers on the same host:
docker-compose up -d && docker-compose up --scale ubuntu14-osquery=20
To stop the containers when running in detached mode like this, use:
docker-compose stop
And to delete the containers when wanting a fresh state, or when finished testing, use:
docker-compose rm
We have had no trouble running up to 100 containerized osqueryd instances on a single processor core and about 1GB of RAM.
Generating a osqueryd core file
The docker containers are configured to allow core files to be generated if osqueryd crashes for some reason. You can attach to the container hosting the errant osqueryd instance, install gdb and use it to read the core file to find out where the crash occurred. The other scenario where you might find a core dump useful is if osqueryd stops responding. In this case you can generate a core dump using the following instructions.
- Open a shell session on a container
docker exec -t -i <container id> /bin/bash
- Find the process ID of osqueryd
ps aux
There will be two osqueryd processes, you'll probably be interested in the child process (the one with the higher pid)
- Send a signal to the process to core dump
kill -3 <pid>
The core file should be in your current working directory on the container.