pm2.config.cjs

/**
 * Configures the process supervisor for Hubble.
 *
 * This ensures if the process fails/dies we recover as quickly as possible.
 * Even if you're using a container process monitor like Docker, some systems
 * (e.g. ECS/Fargate) will still take 1-2 minutes to spin up a new container.
 * Using pm2 allows the container to remain running and thus recover faster.
 */

module.exports = {
  apps: [
    {
      name: "hubble",
      script: "./build/cli.js",
      instances: 1,
      exec_mode: "cluster",
      // HACK: Allows us to pass arguments in a way recognized by the CLI flag processing library we use
      // This is only necessary when running via PM2.
      args: process.env.HUBBLE_ARGS,
      env: {
        "CATCHUP_SYNC_WITH_SNAPSHOT": process.env.CATCHUP_SYNC_WITH_SNAPSHOT ? (process.env.CATCHUP_SYNC_WITH_SNAPSHOT === "true") : "true",
        "HUBBLE_NODE_ARGS": process.env.HUBBLE_NODE_ARGS || "",
      },
      watch: false,
      log_type: "json",
      out_file: "/dev/null",
      err_file: "/dev/stderr",
      restart_delay: 35_000, // Time between restarts is just longer than 30s timeout in hubs
      min_uptime: 60_000, // Min uptime before the app can be considered started
      listen_timeout: 30_000, // Time before forcing a reload if app not listening
      kill_timeout: 60_000, // Time before sending a final SIGKILL after attempting graceful stop
    },
  ],
};