grafana/pkg/services/store/entity
Dan Cech b2b4d89501
Storage: Support grafana.app/folder field selector & label selectors (#79816)
* support grafana.app/folder field selector & label selectors

* lint fix
2023-12-21 18:08:01 -05:00
..
db Storage: make it work for Postgres (#79378) 2023-12-12 18:10:29 +04:00
dummy Storage: Switch from tenant to namespace & remove GRN (#79250) 2023-12-11 12:26:05 -05:00
migrations Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
server Storage: Unified Storage based on Entity API (#71977) 2023-12-06 15:21:21 -05:00
sqlstash Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
tests Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
client_wrapper.go Storage: Switch from tenant to namespace & remove GRN (#79250) 2023-12-11 12:26:05 -05:00
entity_grpc.pb.go Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
entity.pb.go Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
entity.proto Storage: Use int64 for resourceVersion (#79735) 2023-12-20 11:26:35 -08:00
generate.sh Chore: Replace entity GRN with infra/grn GRN (#74198) 2023-08-31 15:43:35 -04:00
key.go Storage: Support grafana.app/folder field selector & label selectors (#79816) 2023-12-21 18:08:01 -05:00
models.go K8s/Folders: Add folders api service (with legacy storage) (#79413) 2023-12-20 20:28:56 +02:00
README.md Chore: Update readme for entity store (#79586) 2023-12-17 13:12:17 +02:00

Unified Storage

The unified storage projects aims to provide a simple and extensible backend to unify the way we store different objects within the Grafana app platform.

It provides generic storage for k8s objects, and can store data either within dedicated tables in the main Grafana database, or in separate storage.

By default it runs in-process within Grafana, but it can also be run as a standalone GRPC service (storage-server).

Storage Overview

There are 2 main tables, the entity table stores a "current" view of the objects, and the entity_history table stores a record of each revision of a given object.

Running Unified Storage

Baseline configuration

The minimum config settings required are:

; dev mode is required
app_mode = development
; need to specify target here for override to work later
target = all

[server]
; https is required for kubectl
protocol = https

[feature_toggles]
; enable unified storage
unifiedStorage = true
; enable k8s apiserver
grafanaAPIServer = true
; store playlists in k8s
kubernetesPlaylists = true
; store json id token in context
idForwarding = true

[grafana-apiserver]
; use unified storage for k8s apiserver
storage_type = unified

With this configuration, you can run everything in-process with:

bra run

The default kubeconfig sends requests directly to the apiserver, to authenticate as a grafana user, create grafana.kubeconfig:

apiVersion: v1
clusters:
- cluster:
    insecure-skip-tls-verify: true
    server: https://127.0.0.1:3000
  name: default-cluster
contexts:
- context:
    cluster: default-cluster
    namespace: default
    user: default
  name: default-context
current-context: default-context
kind: Config
preferences: {}
users:
- name: default
  user:
    username: <username>
    password: <password>

Where <username> and <password> are credentials for basic auth against Grafana.

In this mode, you can interact with the k8s api via:

kubectl --kubeconfig=./grafana.kubeconfig get playlist

To create a playlist, create a file playlist-generate.yaml:

apiVersion: playlist.grafana.app/v0alpha1
kind: Playlist
metadata:
  generateName: x # anything is ok here... except yes or true -- they become boolean!
  labels:
    foo: bar
spec:
  title: Playlist with auto generated UID
  interval: 5m
  items:
  - type: dashboard_by_tag
    value: panel-tests
  - type: dashboard_by_uid
    value: vmie2cmWz # dashboard from devenv

then run:

kubectl --kubeconfig=./grafana.kubeconfig create -f playlist-generate.yaml

Use a separate database

To run against a separate database, update custom.ini:

[entity_api]
db_type = mysql
db_host = localhost:3306
db_name = grafana
db_user = grafanauser
db_pass = grafanapass

MySQL and Postgres are both supported.

Run as a GRPC service

Start GRPC storage-server

This currently only works with a separate database configuration (see previous section).

Start the storage-server with:

GF_DEFAULT_TARGET=storage-server ./bin/grafana server target

The GRPC service will listen on port 10000

Use GRPC server

To run grafana against the storage-server, override the storage_type setting:

GF_GRAFANA_APISERVER_STORAGE_TYPE=unified-grpc ./bin/grafana server

You can then list the previously-created playlists with:

kubectl --kubeconfig=./grafana.kubeconfig get playlist

Changing protobuf interface

  • install protoc
  • install the protocol compiler plugin for Go
go install google.golang.org/grpc/cmd/protoc-gen-go-grpc@latest
  • make changes in .proto file
  • to compile all protobuf files in the repository run make protobuf at its top level