| author | title | summary | tags | ||||||
|---|---|---|---|---|---|---|---|---|---|
Bretton Vine |
Traumadrill |
Traumadrill is a single pot image with multiple benchmarking tools and stress generators for generating artificial system load. |
|
This flavour currently contains stress-ng for generating artificial load and testing alert rules.
It is a work-in-progress for testing prometheus alert rules for system load within a pot environment.
The flavour includes a local consul agent instance to be available that it can connect to (see configuration below). You can e.g. use the consul pot flavour on this site to run consul. You can also connect to this host and service consul restart manually.
- Create your local jail from the image or the flavour files.
- Clone the local jail
- Adjust to your environment:
sudo pot set-env -p <jailname> \ -E DATACENTER=<datacentername> \ -E NODENAME=<nodename> \ -E IP=<IP address of this system> \ -E CONSULSERVERS="<comma-deliminated list of consul servers>" \ -E GOSSIPKEY=<32 byte Base64 key from consul keygen>] \ [ -E REMOTELOG=<IP address> ] - Start the jail
The DATACENTER parameter defines a common datacenter.
The NODENAME parameter defines the name of this node.
The IP parameter is the IP address which will be used to access services.
The CONSULSERVERS parameter is a comma-deliminated list of IP addresses for the consul server or cluster. Do not include spaces!
e.g. CONSULSERVERS="10.0.0.2" or CONSULSERVERS="10.0.0.2,10.0.0.3,10.0.0.4,10.0.0.5,10.0.0.6"
The GOSSIPKEY parameter is the gossip encryption key for consul agent. We're using a default key if you do not set the parameter, do not use the default key for production encryption, instead provide your own.
The REMOTELOG parameter is the IP address of a destination syslog-ng server, such as with the loki flavour, or beast-of-argh flavour.
Open a browser and navigate to http://traumadrill-ip/index.php and click the button to start a test.
Currently a fixed 5min test runs with the following parameters:
stress-ng --cpu 4 --vm 2 --hdd 1 --fork 8 --timeout 5m --metrics-brief --temp-path /tmp/stress-tmp/
There will be no output in the browser until 5min have passed, then brief metrics will be output.
To run another 5min test click the button again.
Watch for prometheus alerts, there should be a high CPU alert, or monitor grafana dashboards to see the load.