helm-charts

mprime (Prime95) Helm Chart

[mprime] is the headless version of Prime95, the software used to search for prime numbers via the Great Internet Mersenne Prime Search (GIMPS).

Installing the Chart

# Make sure you've added this repo to helm
helm repo add brannon https://helm.brannon.online

# Grab the default values, so you can override the default prime.txt config and
# set resource limits
wget -o mprime-value-overrides.yaml https://helm.brannon.online/charts/mprime/values.yaml

# Use your favorite editor to customize the default helm values. Specifically, you'll
# want to replace "ANONYMOUS" with your own username on mersenne.org and set some
# resource limits as mprime is very CPU hungry.
nano mprime-value-overrides.yaml

# Deploy the release using the values you just modified
helm upgrade --install mprime brannon/mprime --values mprime-value-overrides.yaml

Uninstalling the chart

To uninstall/delete the mprime deployment:

helm delete mprime

The command removes all the Kubernetes components associated with the chart and deletes the release.

Generating a `prime.txt` config

mprime is configured via a prime.txt config file located in the same directory as the program itself. This file is generated the first time mprime is run by prompting the user to answer several questions. Because this helm chart is deployed non-interactively, it uses use a pre-configured prime.txt file, the value of which can be found in primeTxtConfig in values.yaml.

If you’d like to generate and use your own prime.txt file, you may do so by following the instructions below.

# Download the existing values.yaml text file and save it as mprime-value-overrides.yaml
# You can ignore this step if you already have a YAML file that defines some overrides
# of the default mprime values.yaml file.
wget -o mprime-value-overrides.yaml https://helm.brannon.online/charts/mprime/values.yaml

# Run mprime inside a docker container to answer the questionnaire it provides
docker run --rm -it --name mprime brannondorsey/mprime

# Once you've completed the questions, open another terminal and run this to
# print the output of prime.txt to the console
docker exec -it mprime cat /prime.txt

# Replace the current value of primeTxtConfig with the output from the command above
nano mprime-value-overrides.yaml

# Be sure the container is removed, I've noticed sometimes mprime doesn't like to shutdown
# nicely
docker rm -f mprime

# Deploy the chart with your own prime.txt values in the primeTxtConfig of mprime-value-overrides.yaml
helm upgrade --install mprime brannon/mprime --values mprime-value-overrides.yaml

Running `mprime` pods at lower priority

This chart provides optional use of PriorityClasses to deprioritize mprime pods in order to prevent scheduling disruptions of more important workloads. Pods running at priority values lower than the default 0 will be preempted (replaced) by more important pods if resource limits become constrained in the cluster. This technique allows you to run lots of mprime workers in your cluster without fear that they will interrupt or prevent production workloads from running: as soon as pods with higher priority are found “Pending”, the scheduler will reap lower priority mprime pods and replace them with the pending pods, if the resources they free would satisfy the pending pods.

This feature is disabled by default, but can be enabled like so:

# Edit via YAML directly in a file passed to --values
priorityClass:
  enabled: true
  value: -1

# Or enable it via the --set argument while installing or upgrading the release
helm upgrade --install --set priorityClass.enabled=true mprime brannon/mprime

Parameters

The following table lists the configurable parameters for this chart and their default values.

Parameter	Description	Default
`replicaCount`	The number of unique mprime instances you’d like to run	`1`
`primeTxtConfig`	The prime.txt configuration to use with each mprime instance. Note that you must replace the entire contents of this property, you cannot override specific values without full replacement.	See values.yaml
`memoryMB`	The max memory mprime will be configured to use, specified in megabytes. A higher value can improve calculation speed. This value is separate from any resource.limits.memory that is set, because it gets translated to a config field in prime.txt. This value MUST be less than any memory limits configured below or pods will receive OOM errors.	`8`
`workersPerReplica`	This configures how many unique workloads are being computed by a replica at any given time. Note that all CPU threads available will be used independent of this value, so setting this value higher won’t necessarily yield faster results (e.g. 1 worker using 4 CPUs may be the same as 2 workers using 2 CPUs).	`2`
`priorityClass.enabled`	Enables the creation of a PriorityClass to be used by pods deployed by the release.	`false`
`priorityClass.value`	The PriorityClass value to use if `priorityClass.enabled` is `true`.	`-1`
`image.repository`	The mprime Docker image	`brannondorsey/mprime`
`image.pullPolicy`	Image pull policy	`IfNotPresent`
`resources`	CPU/Memory resource requests/limits for each mprime replica. It is HIGHLY recommended that you set resource limits as mprime is CPU hungry!	`{}`
`nodeSelector`	Node labels for pod assignment	{}
`tolerations`	Toleration labels for pod assignment	[]
`affinity`	Affinity settings for pod assignment	{}