Chaotic evaluates a plan, how it will bring chaos in your Cloud environment.
Depending on the Cloud API used, it may kill allocations (Hashicorp Nomad), reboot or stop/start virtual machines in your Cloud environment.
With no arguments given, Chaotic runs as a "one shot" meant to be executed as cron job. Passing --periodic runs it as daemon with configurable interval --interval 5 in minutes (1 is the default).
NOTE: The config is re-read on every interval, no need to restart the service after changing the config.
Currently implemented Clouds:
- DigitalOcean
- Vultr
- Hetzner Cloud
- Proxmox KVM
- CloudStack
- Hashicorp Nomad
- cloudscale.ch
pip3 install -U chaotic-ngine
Create a file named config.yaml or use the env var CHAOTIC_CONFIG to point to a config file (also see the example directory):
export CHAOTIC_CONFIG=config_nomad.yaml
Define times when the bot should not doing real actions (it will run in dry-run):
---
kind: ...
excludes:
weekdays:
- Sun
- Sat
times_of_day:
- 22:00-08:00
- 11:00-14:00
days_of_year:
- Jan01
- Apr01
- May01
- Aug01
- Dec24Chaotic will stop a server selected by an optional filter tag and stop/start it with a delay of a configurable time (default 60s).
export CLOUDSTACK_API_KEY="..."
export CLOUDSTACK_API_SECRET="..."
export CLOUDSTACK_API_ENDPOINT="..."
---
kind: cloudstack
dry_run: false
configs:
# Optional, filter tag
tag:
key: chaos
value: enabled
# Optional, 60 seconds is the default
wait_before_restart: 60Chaotic will stop a server selected by an optional filter tag and stop/start it with a delay of a configurable time (default 60s).
export VULTR_API_KEY="..."
---
kind: vultr
dry_run: true
configs:
# Optional instance tag filter
tag: "chaos=opt-in"
# Optional, 60 seconds is the default
wait_before_restart: 60Chaotic will stop a server selected by an optional filter tag and stop/start it with a delay of a configurable time (default 60s).
export CLOUDSCALE_API_TOKEN="..."
---
kind: cloudscale_ch
dry_run: true
configs:
# Optional server tag filter
filter_tag: "chaos=opt-in"
# Optional, 60 seconds is the default
wait_before_restart: 60Chaotic will stop a server selected by an optional filter label and stop/start it with a delay of a configurable time (default 60s).
export HCLOUD_API_TOKEN=...
---
kind: hcloud
dry_run: false
configs:
# Optional server label filter
label: "chaos=enabled"
# Optional, 60 seconds is the default
wait_before_restart: 60Chaotic will stop a droplet selected by an optional filter tag and stop/start it with a delay of a configurable time (default 60s).
export DIGITALOCEAN_ACCESS_TOKEN=...
---
kind: digitalocean
dry_run: false
configs:
# Optional droplet tag filter
tag: "chaos:enabled"
# Optional, 60 seconds is the default
wait_before_restart: 60Chaotic will send an allocation signal to an allocation in the available namespaces selected by an allow list.
export NOMAD_ADDR=http://nomad.example.com:4646
---
kind: nomad
dry_run: true
configs:
experiments:
- job
# Signals to choose from
signals:
- SIGKILL
# Optional: namespace allowlist
namespace_allowlist:
- example-prod
- foobar-prod
# Optional: namespace denylist
namespace_denylist:
- default
# Optional: job type skip list
job_type_skiplist:
- system
- batch
- sysbatch
# Optional: job name skip list
job_skiplist:
- my-job-name
# Optional: Add a meta tag in your nomad job "chaotic" = False to opt-out
job_meta_opt_key: chaoticChaotic will drain a node and set it to be ineligible for some time.
export NOMAD_ADDR=http://nomad.example.com:4646
---
kind: nomad
dry_run: true
configs:
experiments:
- node
# Optional: Node drain deadline in seconds, default 10
node_drain_deadline_seconds: 15
# Optional: Skip nodes in these classes
node_class_skiplist:
- storage
# Optional: Skip nodes with these names
node_skiplist:
- node1
- node5
# Optional: Wait for this amount of seconds before set node to be eligible again, default 60
node_wait_for: 100
# Optional: Also drain system jobs, default false
node_drain_system_jobs: true
# Optional: Drain multiple nodes in one run in percent, fallback 1 node
node_drain_amount_in_percent: 30
Chaotic will stop a VM stop/start it with a delay of a configurable time (default 60s).
export PROXMOX_API_HOST="pve1.example.com"
export PROXMOX_API_USER="root@pam"
export PROXMOX_API_PASSWORD="..."
---
kind: proxmox_kvm
dry_run: false
configs:
# Optional: Do not shutdown VMs having a lower uptime in minutes
min_uptime: 60
# Optional: Do not shutdown VMs in this name list
denylist:
- my-single-vm
# Optional: 60 seconds is the default
wait_before_restart: 60chaos-ngine
One shot:
docker run -ti --rm -v $PWD/examples/config_nomad.yaml:/app/config.yaml -e TZ=Europe/Zurich -e NOMAD_ADDR=$NOMAD_ADDR --name chaotic ghcr.io/ngine-io/chaotic:latest
As service:
docker run -ti --rm -v $PWD/examples/config_nomad.yaml:/app/config.yaml -e TZ=Europe/Zurich -e NOMAD_ADDR=$NOMAD_ADDR --name chaotic ghcr.io/ngine-io/chaotic:latest --periodic
What you should see (e.g. for kind cloudscale.ch):
2021-06-09 09:01:25,433 - cloudscale.log:INFO:Started, version: 0.6.2
2021-06-09 09:01:25,433 - cloudscale.log:INFO:Using profile default
2021-06-09 09:01:25,433 - cloudscale.log:INFO:API Token used: xyz...
2021-06-09 09:01:25,433 - chatic:INFO:Querying with filter_tag: None
2021-06-09 09:01:25,433 - cloudscale.log:INFO:HTTP GET to https://api.cloudscale.ch/v1/servers
2021-06-09 09:01:25,651 - cloudscale.log:INFO:HTTP status code 200
2021-06-09 09:01:25,652 - chatic:INFO:Choose server app3
2021-06-09 09:01:25,653 - chatic:INFO:Stopping server app3
2021-06-09 09:01:25,653 - cloudscale.log:INFO:HTTP POST to https://api.cloudscale.ch/v1/servers/d5628484-a6eb-4ea9-b3ef-ba8da2bb9fe0/stop
2021-06-09 09:01:26,336 - cloudscale.log:INFO:HTTP status code 204
2021-06-09 09:01:26,336 - chatic:INFO:Sleeping for server 60
2021-06-09 09:02:26,393 - cloudscale.log:INFO:HTTP POST to https://api.cloudscale.ch/v1/servers/d5628484-a6eb-4ea9-b3ef-ba8da2bb9fe0/start
2021-06-09 09:02:26,955 - cloudscale.log:INFO:HTTP status code 204
2021-06-09 09:02:26,956 - chatic:INFO:done
![@dependabot[bot]](https://avatars.githubusercontent.com/in/29110?s=64&v=4){kind=small}
