Skip to main content

Frequently Asked Questions (FAQs)

Here are answers to some commonly asked questions. If you have a question that is not covered here, feel free to ask.

Table of Contents​

Do I need a server for launchpad-starter?​

Q: Do I need a server for launchpad-starter?

A: No! The Client Side Tooling that comes with Launchpad should be run on your local machine. These tools are only used to instruct your cluster what to do..

When you setup postgres, how do you configure the zfs storage parameters?​

Q: When you setup postgres, how do you configure the zfs storage parameters (eg the block size, compression, etc) ?

A: Persistent workloads consume Persistent Volumes that use some specific StorageClass (an abstraction). Storage Providers in Kubernetes (like openebs/zfs-localpv), do the operational work of "implementing" those Storage Classes. It is the StorageClass object/resource that would have that particular ZFS setup controlled by parameters. Here's an example of a zfs StorageClass that sets some parameters:

apiVersion: storage.k8s.io/v1
kind: StorageClass
metadata:
  annotations:
    meta.helm.sh/release-name: openebs-zfs-storageclass
    meta.helm.sh/release-namespace: storage
  labels:
    app.kubernetes.io/managed-by: Helm
    launchpad.graphops.xyz/layer: base
    launchpad.graphops.xyz/namespace: storage
  name: openebs-zfs-localpv-compressed-128k
parameters:
  compression: "on"
  fstype: zfs
  poolname: zpool
  recordsize: "128k"
provisioner: zfs.csi.openebs.io
reclaimPolicy: Delete
volumeBindingMode: WaitForFirstConsumer
allowVolumeExpansion: true

Is there a way to inject a pretuned postgres config into the chart?​

Q: Is there a way to inject a pretuned postgres config into the chart? Or is that a post deployment step?

A: Yes. The resource-injector chart allows us to inject a pre-tuned postgres db. This postgresql is a CRD (custom-resource-definition) that is consumed by the postgres-operator (which does the whole operational work of creating the database, setting up users, etc.. as well as replication, backups are also a possiblity). It really allows a great deal of flexibility in terms of what/how it setups the database!

Why are my stateful pods in Pending state and their expected pvc showing WaitForFirstConsumer errors?​

Q: Why are my stateful pods in Pending state and their expected pvc showing WaitForFirstConsumer errors?

  Normal  WaitForPodScheduled   26m (x19 over 31m)   persistentvolume-controller  waiting for pod kube-prometheus-stack-grafana-75b74df8fb-2vwbr to be scheduled
  Normal  WaitForPodScheduled   47s (x102 over 26m)  persistentvolume-controller  waiting for pod kube-prometheus-stack-grafana-75b74df8fb-2vwbr to be scheduled

or 

Normal  WaitForFirstConsumer  6m52s                   persistentvolume-controller  waiting for first consumer to be created before binding

A: Using volumeBindingMode: WaitForFirstConsumer although needed for both openebs-rawfile-localpv and openebs-zfs-localpv seems to misbehave when there is a storageClass set as default in the cluster (the storageClass definition has the following annotation: storageclass.kubernetes.io/is-default-class: "true"). Making sure there is no default storageClass should fix this issue.

Do I need to use Cilium for Launchpad?​

Q: Do I need a specific CNI (Cilium, Calico etc) in order to use Launchpad?

A: The Launchpad stack will work regardless of CNI used and in more general terms should work will all Kubernetes clusters - so you can customize your cluster how you prefer. In our Kubernetes guide we use Cilium due to its use of eBPF technology. This advanced approach offers a significant boost in efficiency, especially noticeable when managing a large number of nodes. It scales well and ensures lower latency, which is crucial for high-performance requirements. While Calico does enjoy a broader base of community support and is a strong choice with its iptables routing, Cilium has the upper advantage due to its performance and its more expansive set of features.

It's important to acknowledge that while Cilium has better performance and features than Calico, it is a bit trickier to set up. Our decision isn't influenced by Launchpad; it's purely a preference based on the operational benefits that Cilium brings to our infrastructure.

Need More Help?​

If your question is not answered here, you can message us on the #kubernetes-launchpad channel on graphprotocol Discord or you can open an issue on our launchpad-namespaces or launchpad-charts repos.