add node provisioning probe

provision-probe/.gitignore

@@ -0,0 +1 @@
+config.yaml

provision-probe/cmd/provision-probe/main.go

@@ -0,0 +1,54 @@
+package main
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"os"
+	"os/signal"
+	"syscall"
+
+	"github.com/rs/zerolog/log"
+	"github.com/threefoldtech/provision-probe/pkg/app"
+	"github.com/threefoldtech/provision-probe/pkg/config"
+	"github.com/threefoldtech/provision-probe/pkg/logger"
+)
+
+func main() {
+	configPath := flag.String("config", "configs/config.yaml", "path to config file")
+	flag.Parse()
+
+	cfg, err := config.Load(*configPath)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "Failed to load config: %v\n", err)
+		os.Exit(1)
+	}
+
+	logger.Init(cfg.LogLevel)
+
+	application, err := app.New(cfg)
+	if err != nil {
+		log.Fatal().Err(err).Msg("Failed to initialize app")
+	}
+	defer application.Close()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	sigChan := make(chan os.Signal, 1)
+	signal.Notify(sigChan, os.Interrupt, syscall.SIGTERM)
+
+	go func() {
+		<-sigChan
+		log.Info().Msg("Shutdown signal received, finishing current work...")
+		cancel()
+
+		<-sigChan
+		log.Warn().Msg("Second shutdown signal received, forcing exit")
+		os.Exit(1)
+	}()
+
+	if err := application.Run(ctx); err != nil {
+		log.Fatal().Err(err).Msg("Service failed")
+	}
+}

provision-probe/configs/config.example.yaml

@@ -0,0 +1,35 @@
+api:
+  host: "0.0.0.0" # host for the API server
+  port: 8080 # port for the API server
+
+timescaledb:
+  url: "postgresql://postgres:postgres@localhost:5432/provision_probe" # url for the timescaledb database
+
+
+probe: 
+  interval: "6h" # interval for the deployment cycle
+  concurrency_limit: 10 # concurrency limit for the deployment cycle
+  timeout: "10m" # timeout for the deployment cycle
+  workload_size: "light" # light, medium, heavy
+  shutdown_timeout: "30s" # timeout for graceful shutdown
+  retry:
+    max_retries: 3 # maximum number of retries for transient failures
+    initial_backoff: "1s" # initial backoff duration between retries
+    max_backoff: "30s" # maximum backoff duration
+    multiplier: 2.0 # exponential backoff multiplier 
+
+scoring: 
+  window: "90d" # default window for the scoring (overwritten by the window query parameter)
+  min_attempts: 1 # default minimum attempts for the scoring (overwritten by the min_attempts query parameter)
+
+grid:
+  network: "dev" # dev, qa, test, main
+  mnemonic: "" # mnemonic for the grid
+
+nodes:
+  status: "up" # up, healthy
+  farms: [] # farm IDs to include (empty means all farms)
+  nodes: [] # node IDs to include (empty means all nodes)
+  exclude: [] # node IDs to exclude
+
+log_level: "info"

provision-probe/docker-compose.yml

@@ -0,0 +1,30 @@
+services:
+  timescaledb:
+    image: timescale/timescaledb:latest-pg16
+    container_name: provision-probe-db
+    environment:
+      POSTGRES_USER: postgres
+      POSTGRES_PASSWORD: postgres
+      POSTGRES_DB: provision_probe
+    ports:
+      - "${PG_PORT:-5432}:5432"
+    volumes:
+      - timescaledb-data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U postgres"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  pgweb:
+    image: sosedoff/pgweb:latest
+    container_name: provision-probe-pgweb
+    ports:
+      - "${VIEW_PORT:-8081}:8081"
+    environment:
+      - PGWEB_DATABASE_URL=postgres://postgres:postgres@timescaledb:5432/provision_probe?sslmode=disable
+    depends_on:
+      - timescaledb
+
+volumes:
+  timescaledb-data: