seaweedfs-k8s-operator

Purpose

This charm is intended to be used as a stand-in for s3-integrator + (micro)ceph for testing purposes. When testing charms, instead of deploying (micro)ceph, s3-integrator, setting config options and running the sync-credentials action, this charm mimics the s3 relation interface and provides an s3 endpoint.

Compared to s3-integrator

Relation data

This charm does not use the s3 library. Instead, it renderes the relation data itself. For a related charm, relation data may look like this:

  - relation-id: 7
    endpoint: s3
    related-endpoint: s3-credentials
    application-data:
      access-key: placeholder
      bucket: s3-credentials-7
      endpoint: http://swfs-0.swfs-endpoints.welcome-k8s.svc.cluster.local:8333
      secret-key: placeholder

Note that the bucket name is automatically derived from the relation id. No config options needed.

Bucket name

In the past there has been some confusion about who decides on the bucket name - the requesting charm (e.g. mimir, loki, tempo), or the s3-integrator charm. It seems like everyone agrees now that it's the s3-integrator where the bucket name should be set (via config option).

In this charm, the same principle holds, but there is no config option for bucket name, because:

1. For testing purposes, we don't care that the bucket name is not fixed.
2. This way we could relate multiple charms to the same seaweedfs charm, unlike s3-integrator where each app (mimir, loki, tempo) have their own s3-integrator due to different bucket names.

Usage example

Here's a sample bundle

graph LR
mc ---|s3:s3-credentials| swfs
mw ---|mimir-cluster| mc
tw ---|tempo-cluster| tc
tc ---|s3:s3-credentials| swfs

Loading

bundle: kubernetes
applications:
  mc:
    charm: mimir-coordinator-k8s
    channel: 1/edge
    revision: 43
    base: ubuntu@22.04/stable
    resources:
      nginx-image: 14
      nginx-prometheus-exporter-image: 4
    scale: 1
    constraints: arch=amd64
  mw:
    charm: mimir-worker-k8s
    channel: 1/edge
    revision: 50
    base: ubuntu@22.04/stable
    resources:
      mimir-image: 16
    scale: 1
    options:
      role-all: true
    constraints: arch=amd64
    trust: true
  swfs:
    charm: seaweedfs-k8s
    channel: edge
    revision: 5
    base: ubuntu@24.04/stable
    scale: 1
    constraints: arch=amd64
  tc:
    charm: tempo-coordinator-k8s
    channel: 1/edge
    revision: 79
    base: ubuntu@22.04/stable
    resources:
      nginx-image: 7
      nginx-prometheus-exporter-image: 4
    scale: 1
    constraints: arch=amd64
    trust: true
  tw:
    charm: tempo-worker-k8s
    channel: 1/edge
    revision: 59
    base: ubuntu@22.04/stable
    resources:
      tempo-image: 6
    scale: 1
    options:
      role-all: true
    constraints: arch=amd64
    trust: true
relations:
- - mc:s3
  - swfs:s3-credentials
- - mw:mimir-cluster
  - mc:mimir-cluster
- - tw:tempo-cluster
  - tc:tempo-cluster
- - tc:s3
  - swfs:s3-credentials

Manual testing

juju ssh --container seaweedfs swfs/0 /charm/bin/pebble logs -f | grep -iE "error|fail"

# Make sure size is growing
juju ssh --container seaweedfs swfs/0 du -hc /data

# Refs:
# https://github.com/seaweedfs/seaweedfs/blob/master/docker/compose/local-filer-backup-compose.yml
# https://github.com/seaweedfs/seaweedfs/blob/master/.github/workflows/s3tests.yml

UNIT=$(juju status --format=yaml | yq '.applications.swfs.units.swfs/0.address')

curl --fail -I http://$UNIT:9333/cluster/healthz

# Master server
curl -s http://$UNIT:9333/cluster/status

# Volume server
curl -s http://$UNIT:8080/status

# Filer
curl -s http://$UNIT:8888/

# S3 server
curl -s http://$UNIT:8333/

sudo apt install s3cmd

s3cmd --host=$UNIT:8333 --access_key=placeholder --secret_key=placeholder --host-bucket= --no-ssl ls
s3cmd --host=$UNIT:8333 --access_key=placeholder --secret_key=placeholder --host-bucket= --no-ssl mb s3://loki