parca-dev/helm-charts

Kubernetes Helm Charts to run Parca and Parca Agent

helm-charts's Issues

Move default branch to main

I suggest to move the default branch for this repository to main to be in line with the other parca repositories and current community defaults.

GitHub allows to do that in the repository settings with just a few mouse clicks and automatically re-targets PRs to the new default branch.

Add StatefulSet setup to persist data

The chart should have the option to be used as a StatefulSet and --enable-persistence

move to templated helm docs

Current docs are manually generated with helm-docs (https://github.com/norwoodj/helm-docs) from values.yaml
Better to provide template where all dynamic stuff will be generated from current variables (such as appversion and etc)
Grafana have nice tidy example https://github.com/grafana/helm-charts/blob/main/charts/tempo/README.md.gotmpl

[proposal] split charts for agent and server

As the two components might be deployed in different environment/cluster configurations, imo it makes sense to have separate charts to avoid cluttering them with options.
It is just a suggestion, I can refactor them in a PR if it makes sense for others as well

helm chart doesn't deploy

I think the healthcheck code changed in this commit

Configuration file for server doesn't allow unexpected values

I'm attempting to use this chart with the latest server, and it is expecting different values in the configuration file for the server.

It is currently impossible to use with the current chart as it is difficult to clear maps: helm/helm#5184

I will add ability to fully override the contents of the configuration file

add github action check for helm-docs

The github pr CI should detect whether the readme has been templatesd according to the readme.md.tpl file, and fail the build if not matching

Add the ability to use a Parca Server deployed in a different cluster

Thanks for this chart.

Currently the chart is assuming that the agents and server will be deployed in the same Kubernetes cluster. In some setup, we might see servers deploy in a separate cluster, in a centralized fashion.

Are you open to support this use case ?

agent: Socket path should be left empty by default

The charts specify socketPath flag by default and the value is for containerd runtime (/run/containerd/containerd.sock). Unfortunately, this prevents us from inferring the correct container runtime for the agent. This should be left blank by default.

helm-charts/charts/parca/templates/agent-daemonset.yaml

Line 44 in b2756c0

- --socket-path={{ .Values.agent.socketPath }}

helm-charts/charts/parca/values.yaml

Line 19 in 6798d97

socketPath: /run/containerd/containerd.sock

Parca Agent: Allow configuring remaining `--remote*` flags

Implement these missing flags:

• --remote-store-bearer-token=STRING
• --remote-store-bearer-token-file=STRING
• --remote-store-insecure
• --remote-store-insecure-skip-verify
• --remote-store-debuginfo-upload-disable
• --remote-store-batch-write-interval=10s