underpost 3.2.30 → 3.2.80

package/src/api/file/file.router.js

@@ -1,12 +1,10 @@
+import express from 'express';
 import { adminGuard } from '../../server/auth.js';
-import { loggerFactory } from '../../server/logger.js';
 import { FileController } from './file.controller.js';
-import express from 'express';
-
-const logger = loggerFactory(import.meta);
 
 class FileRouter {
   /**
+   * authenticated writes, admin-only deletes.
    * @param {import('../types.js').RouterOptions} options
    * @returns {import('express').Router}
    */
@@ -17,8 +15,18 @@ class FileRouter {
     router.get(`/blob/:id`, async (req, res) => await FileController.get(req, res, options));
     router.get(`/:id`, async (req, res) => await FileController.get(req, res, options));
     router.get(`/`, async (req, res) => await FileController.get(req, res, options));
-    router.delete(`/:id`, options.authMiddleware, adminGuard, async (req, res) => await FileController.delete(req, res, options));
-    router.delete(`/`, options.authMiddleware, adminGuard, async (req, res) => await FileController.delete(req, res, options));
+    router.delete(
+      `/:id`,
+      options.authMiddleware,
+      adminGuard,
+      async (req, res) => await FileController.delete(req, res, options),
+    );
+    router.delete(
+      `/`,
+      options.authMiddleware,
+      adminGuard,
+      async (req, res) => await FileController.delete(req, res, options),
+    );
     return router;
   }
 }

package/src/api/test/test.controller.js

@@ -1,59 +1,14 @@
-import { loggerFactory } from '../../server/logger.js';
+import { controllerHandler, sendSuccess, serviceHandler } from '../../server/middlewares.js';
 import { TestService } from './test.service.js';
 
-const logger = loggerFactory(import.meta);
-
 class TestController {
-  static post = async (req, res, options) => {
-    try {
-      return res.status(200).json({
-        status: 'success',
-        data: await TestService.post(req, res, options),
-      });
-    } catch (error) {
-      logger.error(error, error.stack);
-      return res.status(400).json({
-        status: 'error',
-        message: error.message,
-      });
-    }
-  };
-  static get = async (req, res, options) => {
-    try {
-      const result = await TestService.get(req, res, options);
-      if (result)
-        return res.status(200).json({
-          status: 'success',
-          data: result,
-        });
-      else
-        return res.status(400).json({
-          status: 'error',
-        });
-    } catch (error) {
-      logger.error(error, error.stack);
-      return res.status(400).json({
-        status: 'error',
-        message: error.message,
-      });
-    }
-  };
-  static delete = async (req, res, options) => {
-    try {
-      const result = await TestService.delete(req, res, options);
-
-      return res.status(200).json({
-        status: 'success',
-        data: result,
-      });
-    } catch (error) {
-      logger.error(error, error.stack);
-      return res.status(400).json({
-        status: 'error',
-        message: error.message,
-      });
-    }
-  };
+  static post = serviceHandler(TestService.post);
+  static get = controllerHandler(async (req, res, options) => {
+    const result = await TestService.get(req, res, options);
+    if (result) return sendSuccess(res, result);
+    return res.status(400).json({ status: 'error' });
+  });
+  static delete = serviceHandler(TestService.delete);
 }
 
 export { TestController };

package/src/api/test/test.router.js

@@ -1,8 +1,5 @@
-import { loggerFactory } from '../../server/logger.js';
-import { TestController } from './test.controller.js';
 import express from 'express';
-
-const logger = loggerFactory(import.meta);
+import { TestController } from './test.controller.js';
 
 class TestRouter {
   /**

package/src/cli/cluster.js

@@ -876,6 +876,73 @@ net.ipv4.ip_forward = 1' | sudo tee /etc/sysctl.d/99-k3s.conf > /dev/null`,
       }
     },
 
+    /**
+     * @method _pruneContainerCaches
+     * @description Reclaims container-runtime disk left behind after a cluster
+     * teardown: stopped containers, unused images, build cache and anonymous
+     * volumes. Best-effort across every runtime present on the host (docker,
+     * podman, and optionally the CRI runtime via crictl). Each block is guarded
+     * by a command-existence check, so it is safe to call when a runtime is
+     * absent. This is what turns a "cluster deleted" into "disk actually freed":
+     * `kind delete` / `kubeadm reset` remove the cluster but leave gigabytes of
+     * images and overlay layers under /var/lib/{docker,containers}.
+     * @param {object} [options]
+     * @param {boolean} [options.all=true] - Remove all unused images, not just dangling ones.
+     * @param {boolean} [options.crictl=false] - Also prune the CRI runtime via crictl.
+     * @param {string} [options.criSocket] - Optional crictl --runtime-endpoint override.
+     * @private
+     */
+    _pruneContainerCaches(options = {}) {
+      const all = options.all !== false;
+      const a = all ? '-a ' : '';
+      logger.info(`  -> Pruning container-runtime caches (all=${all})...`);
+      // Docker (also matches the podman-docker shim when docker is symlinked to podman).
+      shellExec(
+        `if command -v docker >/dev/null 2>&1; then sudo docker system prune ${a}--volumes -f; sudo docker builder prune ${a}-f; fi`,
+        { silentOnError: true },
+      );
+      // Podman native — on this host images/overlays live under /var/lib/containers/storage.
+      shellExec(`if command -v podman >/dev/null 2>&1; then sudo podman system prune ${a}--volumes -f; fi`, {
+        silentOnError: true,
+      });
+      if (options.crictl) {
+        const ep = options.criSocket ? `--runtime-endpoint ${options.criSocket} ` : '';
+        shellExec(
+          `if command -v crictl >/dev/null 2>&1; then sudo env PATH="$PATH:/usr/local/bin:/usr/bin" crictl ${ep}rmi --prune; fi`,
+          { silentOnError: true },
+        );
+      }
+      Underpost.cluster._unmountOrphanContainerOverlays();
+    },
+
+    /**
+     * @method _unmountOrphanContainerOverlays
+     * @description Unmounts leaked container overlay 'merged' mounts under
+     * /var/lib/{containers/storage,docker}/overlay. Pruning images/containers
+     * frees the backing data but leaves these mountpoints attached, so they
+     * keep showing up as identical `overlay ... /merged` rows in `df -h` long
+     * after the containers are gone. Only overlays NOT backing a still-existing
+     * container are unmounted, so running workloads are never disturbed.
+     * Best-effort; safe to call when no runtime is present.
+     * @private
+     */
+    _unmountOrphanContainerOverlays() {
+      logger.info('  -> Unmounting orphaned container overlay mounts...');
+      shellExec(`if command -v podman >/dev/null 2>&1; then sudo podman umount --all >/dev/null 2>&1 || true; fi`, {
+        silentOnError: true,
+      });
+      shellExec(
+        `active="$(sudo podman ps -aq 2>/dev/null | xargs -r -I{} sudo podman inspect --format '{{.GraphDriver.Data.MergedDir}}' {} 2>/dev/null || true)"
+findmnt -rn -o TARGET 2>/dev/null | grep -E '/var/lib/(containers/storage|docker)/overlay.*/merged' | sort -r | while IFS= read -r m; do
+  if ! printf '%s\\n' "$active" | grep -qxF "$m"; then
+    echo "Unmounting orphaned overlay: $m"
+    sudo umount -l "$m" 2>/dev/null || true
+  fi
+done`,
+        { silentOnError: true },
+      );
+    },
+
     /**
      * @method _lazyUmountKubeletMounts
      * @description Lazy-unmounts every mount under /var/lib/kubelet so a
@@ -910,7 +977,7 @@ net.ipv4.ip_forward = 1' | sudo tee /etc/sysctl.d/99-k3s.conf > /dev/null`,
       if (options.removeVolumeHostPaths) Underpost.cluster._cleanHostPathPvs();
       else logger.info('  -> Skipping (pass --remove-volume-host-paths to enable).');
 
-      logger.info('Phase 3/5: Deleting all Kind clusters...');
+      logger.info('Phase 3/6: Deleting all Kind clusters...');
       shellExec(`clusters=$(kind get clusters)
 if [ -n "$clusters" ]; then
   for c in $clusters; do
@@ -919,11 +986,14 @@ if [ -n "$clusters" ]; then
   done
 fi`);
 
-      logger.info('Phase 4/5: Cleaning kubeconfig and Kind Docker networks...');
+      logger.info('Phase 4/6: Cleaning kubeconfig and Kind Docker networks...');
       shellExec(`rm -rf "$HOME/.kube"`);
       Underpost.cluster.recoverKindDockerNetworks();
 
-      logger.info('Phase 5/5: Re-applying host configuration (Docker, containerd, sysctl).');
+      logger.info('Phase 5/6: Pruning container-runtime caches (kindest/node images, build cache, volumes)...');
+      Underpost.cluster._pruneContainerCaches({ all: true });
+
+      logger.info('Phase 6/6: Re-applying host configuration (Docker, containerd, sysctl).');
       Underpost.cluster.config();
 
       logger.info('=== KIND SAFE RESET COMPLETE ===');
@@ -995,7 +1065,7 @@ fi`);
       shellExec(`if ip link show tunl0 >/dev/null 2>&1; then sudo ip link del tunl0; fi`);
       shellExec(`sudo iptables -F`);
       shellExec(`sudo iptables -t nat -F`);
-      shellExec(`if command -v crictl >/dev/null 2>&1; then sudo crictl rmi --prune; fi`);
+      Underpost.cluster._pruneContainerCaches({ all: true, crictl: true });
 
       logger.info('Phase 7/7: Re-applying host configuration (Docker, containerd, sysctl).');
       Underpost.cluster.config();
@@ -1042,6 +1112,8 @@ fi`);
       shellExec(`if [ -d /etc/rancher/k3s ]; then sudo rm -rf /etc/rancher/k3s; fi`);
       shellExec(`if ip link show flannel.1 >/dev/null 2>&1; then sudo ip link del flannel.1; fi`);
       shellExec(`if ip link show cni0 >/dev/null 2>&1; then sudo ip link del cni0; fi`);
+      // k3s-uninstall.sh removes /var/lib/rancher/k3s; still prune any host docker/podman leftovers.
+      Underpost.cluster._pruneContainerCaches({ all: true });
 
       logger.info('Phase 5/5: Re-applying minimal K3s host config.');
       Underpost.cluster.configMinimalK3s();
@@ -1125,13 +1197,34 @@ exclude=kubelet kubeadm kubectl cri-tools kubernetes-cni
 EOF`);
       shellExec(`sudo yum install -y kubelet kubeadm kubectl --disableexcludes=kubernetes`);
 
-      // Install Helm
-      shellExec(`curl -fsSL -o get_helm.sh https://cdn.jsdelivr.net/gh/helm/helm@main/scripts/get-helm-3`);
-      shellExec(`chmod 700 get_helm.sh`);
-      shellExec(`./get_helm.sh`);
-      shellExec(`chmod +x /usr/local/bin/helm`);
-      shellExec(`sudo mv /usr/local/bin/helm /bin/helm`);
-      shellExec(`sudo rm -rf get_helm.sh`);
+      // Install Helm — check if already installed and linked to /bin/helm first
+      const helmBin = shellExec(`test -x /bin/helm && echo exists || echo missing`, {
+        stdout: true,
+        silent: true,
+      }).trim();
+      const helmLocalBin = shellExec(`test -x /usr/local/bin/helm && echo exists || echo missing`, {
+        stdout: true,
+        silent: true,
+      }).trim();
+
+      if (helmBin === 'exists') {
+        logger.info('Helm is already installed and linked to /bin/helm; skipping.');
+      } else if (helmLocalBin === 'exists') {
+        // Helm binary exists at /usr/local/bin but not linked to /bin/helm
+        logger.info('Helm found at /usr/local/bin; linking to /bin/helm...');
+        shellExec(`sudo ln -sf /usr/local/bin/helm /bin/helm`);
+      } else {
+        // Helm not installed — download and install
+        shellExec(`curl -fsSL -o get_helm.sh https://cdn.jsdelivr.net/gh/helm/helm@main/scripts/get-helm-3`);
+        shellExec(`chmod 700 get_helm.sh`);
+        // Run get_helm.sh but ignore its exit code — it may fail on PATH check
+        // even when installation succeeds (binary placed at /usr/local/bin).
+        shellExec(`bash ./get_helm.sh || true`);
+        // Ensure the binary is executable and linked to /bin/helm
+        shellExec(`test -x /usr/local/bin/helm && sudo chmod +x /usr/local/bin/helm || true`);
+        shellExec(`test -x /usr/local/bin/helm && sudo ln -sf /usr/local/bin/helm /bin/helm || true`);
+        shellExec(`sudo rm -rf get_helm.sh`);
+      }
 
       // Install snap
       shellExec(`sudo yum install -y snapd`);

package/src/cli/db.js

@@ -6,7 +6,7 @@
  * Supports MariaDB and MongoDB with import/export capabilities, Git integration, and multi-pod operations.
  */
 
-import { mergeFile, splitFileFactory, loadConfServerJson, resolveConfSecrets } from '../server/conf.js';
+import { mergeFile, splitFileFactory, loadConfServerJson, resolveConfSecrets, loadConfInstances } from '../server/conf.js';
 import { loggerFactory } from '../server/logger.js';
 import { shellExec } from '../server/process.js';
 import fs from 'fs-extra';
@@ -1119,7 +1119,9 @@ class UnderpostDB {
             // Process additional instances
             const confInstancesPath = `./engine-private/conf/${deployId}/conf.instances.json`;
             if (fs.existsSync(confInstancesPath)) {
-              const confInstances = JSON.parse(fs.readFileSync(confInstancesPath, 'utf8'));
+              // Expands multiInstance variants so each deployed instance is
+              // registered (handles both the bare-array and { instances } shapes).
+              const confInstances = loadConfInstances(deployId);
               for (const instance of confInstances) {
                 const { id, host, path, fromPort, metadata } = instance;
                 const { runtime } = metadata;

package/src/cli/deploy.js

@@ -12,6 +12,7 @@ import {
   cronDeployIdResolve,
   deployRangePortFactory,
   getDataDeploy,
+  loadConfInstances,
   loadConfServerJson,
   loadReplicas,
   pathPortAssignmentFactory,
@@ -22,8 +23,25 @@ import { INTERNAL_READY_PATH, INTERNAL_HEALTH_PATH } from '../server/runtime-sta
 import fs from 'fs-extra';
 import dotenv from 'dotenv';
 import os from 'node:os';
+import crypto from 'node:crypto';
 import Underpost from '../index.js';
 
+/**
+ * Clamps an identifier to the Kubernetes DNS-1123 label limit (63 chars),
+ * used for pod-local `volumes[].name` / `volumeMounts[].name`. Names within the
+ * limit are returned verbatim so existing short names are stable; longer ones
+ * are truncated and suffixed with an 8-char content hash to stay unique and
+ * deterministic (e.g. the per-variant instance volume names, which append the
+ * full `<deployId>-<env>-<traffic>` and can exceed 63).
+ * @param {string} name - Candidate name.
+ * @returns {string} A name no longer than 63 characters.
+ */
+const k8sVolumeName = (name) => {
+  if (typeof name !== 'string' || name.length <= 63) return name;
+  const hash = crypto.createHash('sha1').update(name).digest('hex').slice(0, 8);
+  return `${name.slice(0, 54)}-${hash}`;
+};
+
 const logger = loggerFactory(import.meta);
 
 /**
@@ -562,7 +580,7 @@ ${Underpost.deploy
           const yamlPath = `./engine-private/conf/${deployId}/build/${env}/secret.yaml`;
           fs.writeFileSync(yamlPath, secretYaml, 'utf8');
         } else {
-          const deploymentsFiles = ['Dockerfile', 'proxy.yaml', 'deployment.yaml', 'pv-pvc.yaml'];
+          const deploymentsFiles = ['Dockerfile', 'proxy.yaml', 'deployment.yaml', 'pv-pvc.yaml', 'grpc-service.yaml'];
           for (const file of deploymentsFiles) {
             if (fs.existsSync(`./engine-private/conf/${deployId}/build/${env}/${file}`)) {
               fs.copyFileSync(
@@ -668,7 +686,7 @@ spec:
      * @returns {string|null} - Current traffic status ('blue' or 'green') or null if not found.
      * @memberof UnderpostDeploy
      */
-    getCurrentTraffic(deployId, options = { hostTest: '', namespace: '' }) {
+    getCurrentTraffic(deployId, options = { hostTest: '', namespace: '', env: '' }) {
       if (!options.namespace) options.namespace = 'default';
       // kubectl get deploy,sts,svc,configmap,secret -n default -o yaml --export > default.yaml
       const hostTest = options?.hostTest
@@ -683,6 +701,18 @@ spec:
         silentOnError: true,
       });
       if (!info) return null;
+      // Env-scoped resolution: read THIS deploy's colour from its own service
+      // name (`<deployId>-<env>-<colour>-service`). Essential for shared
+      // multi-instance hosts, where one HTTPProxy holds several variants' routes
+      // on possibly different colours — a whole-document `.match('blue')` would
+      // return whichever colour appears first, i.e. a sibling's, not this one's.
+      // The regex is anchored on the full `<deployId>-<env>-` prefix so
+      // `dd-cyberia-mmo-server` never matches `dd-cyberia-mmo-server-forest`.
+      if (options.env) {
+        const escaped = deployId.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+        const match = info.match(new RegExp(`${escaped}-${options.env}-(blue|green)-service`));
+        return match ? match[1] : null;
+      }
       return info.match('blue') ? 'blue' : info.match('green') ? 'green' : null;
     },
 
@@ -825,9 +855,9 @@ EOF`);
           const deployId = _deployId.trim();
           const instances = [];
           if (fs.existsSync(`./engine-private/conf/${deployId}/conf.instances.json`)) {
-            const confInstances = JSON.parse(
-              fs.readFileSync(`./engine-private/conf/${deployId}/conf.instances.json`, 'utf8'),
-            );
+            // Expands multiInstance variants so status lists every deployed
+            // instance (mmo-server, mmo-server-forest, …), not just the templates.
+            const confInstances = loadConfInstances(deployId);
             for (const instance of confInstances) {
               const _deployId = `${deployId}-${instance.id}`;
               instances.push({
@@ -838,7 +868,7 @@ EOF`);
                 toPort: instance.toPort,
                 fromDebugPort: instance.fromDebugPort,
                 toDebugPort: instance.toDebugPort,
-                traffic: Underpost.deploy.getCurrentTraffic(_deployId, { namespace, hostTest: instance.host }),
+                traffic: Underpost.deploy.getCurrentTraffic(_deployId, { namespace, hostTest: instance.host, env }),
               });
             }
           }
@@ -1009,10 +1039,17 @@ EOF`);
      */
     configMap(env, namespace = 'default') {
       const cronDeployId = cronDeployIdResolve() || 'dd-cron';
+      const envFilePath = `/home/dd/engine/engine-private/conf/${cronDeployId}/.env.${env}`;
+      // `--from-env-file` turns every KEY=VALUE into a secret key that the Deployment injects via
+      // `envFrom`. Strip shell/runtime-critical keys (notably PATH) first — an injected PATH
+      // overrides the image's own and breaks coreutils/sudo resolution inside the pod.
+      const sanitizedEnvPath = `${envFilePath}.secret`;
+      fs.writeFileSync(sanitizedEnvPath, Underpost.secret.sanitizeSecretEnvFile(fs.readFileSync(envFilePath, 'utf8')));
       shellExec(`kubectl delete secret underpost-config -n ${namespace} --ignore-not-found`);
       shellExec(
-        `kubectl create secret generic underpost-config --from-env-file=/home/dd/engine/engine-private/conf/${cronDeployId}/.env.${env} --dry-run=client -o yaml | kubectl apply -f - -n ${namespace}`,
+        `kubectl create secret generic underpost-config --from-env-file=${sanitizedEnvPath} --dry-run=client -o yaml | kubectl apply -f - -n ${namespace}`,
       );
+      fs.removeSync(sanitizedEnvPath);
     },
     /**
      * Switches the traffic for a deployment.
@@ -1068,7 +1105,13 @@ EOF`);
      *      runners derive it from the comma-path field or `--node-name`
      *      (`run sync`: `path.split(',')[4]` > `--node-name` > default) and from
      *      `--node-name` directly (`run instance`).
-     *   2. **Cluster-type default** — when no explicit node is given: `kind-worker`
+     *   2. **`UNDERPOST_DEPLOY_NODE` env** — for kubeadm / k3s, the configured
+     *      target node name. This makes hostPath PV `nodeAffinity` deterministic
+     *      regardless of where the manifest is *built*: building inside a
+     *      container or CI runner would otherwise leak that box's `os.hostname()`
+     *      (e.g. a random container id) into `nodeSelector`, pinning the PV to a
+     *      node that does not exist in the cluster.
+     *   3. **Cluster-type default** — when nothing above is set: `kind-worker`
      *      for a kind cluster (the node that hosts kind hostPath volumes),
      *      otherwise the control-plane / current host (`os.hostname()`) for
      *      kubeadm / k3s. With no explicit cluster flag, `development` is treated
@@ -1086,7 +1129,8 @@ EOF`);
     resolveDeployNode({ node = '', kind = false, kubeadm = false, k3s = false, env = '' } = {}) {
       if (node) return node;
       const isKind = kind || (!kubeadm && !k3s && env !== 'production');
-      return isKind ? 'kind-worker' : os.hostname();
+      if (isKind) return 'kind-worker';
+      return process.env.UNDERPOST_DEPLOY_NODE || os.hostname();
     },
 
     /**
@@ -1234,13 +1278,18 @@ EOF
           volumeName = `${volumeName}-${version}`;
           claimName = claimName ? `${claimName}-${version}` : null;
         }
+        // The pod-local volume name is a DNS-1123 label (max 63 chars); the PVC
+        // `claimName` it references is a subdomain (max 253) and stays verbatim.
+        // Per-variant instance names append <deployId>-<env>-<traffic> and can
+        // exceed 63, so clamp only the pod-local name (mount name must match it).
+        const podVolumeName = k8sVolumeName(volumeName);
         _volumeMounts += `
-        - name: ${volumeName}
+        - name: ${podVolumeName}
           mountPath: ${volumeMountPath}
 ${secret ? `          readOnly: true\n` : ''}`;
 
         _volumes += `
-    - name: ${volumeName}
+    - name: ${podVolumeName}
  ${
    emptyDir
      ? `     emptyDir: {}`