update default values.yaml

[thobianchi]

Signed-off-by: Thomas Bianchi [email protected]

What this PR does:
edit values.yaml giving a working cortex instance configured with block storage, memcache and memberlist

Which issue(s) this PR fixes:
[N/A]

Checklist

• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

This is a proposal to update values.yaml giving a working instance with updated defaults.
I tested with a kubernetes kind created simply by: kind create cluster and then helm install --create-namespace --namespace cortex -f values.yaml cortex ., eventually every component start:

in depth:

• I configured a backend blocks by default on filesystem, we are using in our production environment gcp bucket storage, so I included the example for it ( it needs only two fields name and json ).

  blocks-storage-memcached: true

  storage:
    engine: blocks
  blocks_storage:
    backend: "filesystem"
    filesystem:
      dir: "/data/blocks-storage"
    # backend: "gcs"
    # gcs:
    #   bucket_name: ""
    #   service_account: "" # json credentials
    bucket_store:
      sync_dir: "data"
      index_cache:
        backend: memcached

• The numbers in this section are taken partially from this issue. I found that these defaults are better for a real environment without being too much.

       memcached:
          addresses: 'dns+{{ include "cortex.fullname" $ }}-memcached-blocks-index:11211'
          timeout: 300ms
          max_idle_connections: 750
          max_async_concurrency: 100
          max_async_buffer_size: 10000000

• Memberlist configuration instead of consul

    ring:
      kvstore:
        store: memberlist

• Decreased all components to 1 replica
• Memcached Chart do not have those values like extraArgs or maxItemMemory so I found out that this works:

memcached-frontend:
  enabled: true
  architecture: "high-availability"
  replicaCount: 1
  # dpdbMinAvailable: 1
  # image: memcached:1.5.7-alpine
  command:
    - /run.sh
    - -m 256
    - -I 32m
    - -t 4
  arguments: []
  resources: {}
  #  requests:
  #    memory: 256Mi
  #    cpu: 250m
  #  limits:
  #    memory: 256Mi
  #    cpu: 250m
  metrics:
    enabled: false
    serviceMonitor:
      enabled: false

We have been using Cortex in production environment for 8 months, with great results! I want to thank you all.
Our configuration is really similar to this proposal.

Some numbers:

I think would be useful insert in this repository a directory cortex-mixin with a collection of Grafana dashboard for monitoring cortex itself, I do not even remember where I found this dashboard...

[nschad]

I think would be useful insert in this repository a directory cortex-mixin with a collection of Grafana dashboard for monitoring cortex itself, I do not even remember where I found this dashboard...

You probably mean this: https://github.com/grafana/cortex-jsonnet

This PR is very similiar with #227. Maybe check that PR first and write down your thoughts as comments?

[thobianchi]

oh.. I missed that PR, yes I will look into it if I can contribute there.