pikasTech-HWLAB/scripts/ci/runtime-ready.mjs

#!/usr/bin/env node
// Wait for Argo/runtime convergence with a single wall-clock budget.

import { readFileSync } from "node:fs";
import https from "node:https";

const host = process.env.KUBERNETES_SERVICE_HOST;
const port = process.env.KUBERNETES_SERVICE_PORT || "443";
const startedAt = Date.now();
const timeoutMs = positiveInteger(process.env.HWLAB_RUNTIME_READY_TIMEOUT_MS, 60_000);
const pollMs = positiveInteger(process.env.HWLAB_RUNTIME_READY_POLL_MS, 1_000);
const progressEveryMs = positiveInteger(process.env.HWLAB_RUNTIME_READY_PROGRESS_MS, 10_000);
const deadline = startedAt + timeoutMs;

function requiredEnv(name) {
  const value = process.env[name];
  if (!value) throw new Error(name + " is required");
  return value;
}

const revision = requiredEnv("HWLAB_SOURCE_REVISION");
const runtimeNamespace = requiredEnv("HWLAB_RUNTIME_NAMESPACE");
const gitopsTarget = requiredEnv("HWLAB_GITOPS_TARGET");
const pipelineRun = process.env.HWLAB_TEKTON_PIPELINERUN || null;
const taskRun = process.env.HWLAB_TEKTON_TASKRUN || null;
const task = process.env.HWLAB_TEKTON_TASK || null;

function positiveInteger(value, fallback) {
  const parsed = Number.parseInt(String(value ?? ""), 10);
  return Number.isFinite(parsed) && parsed > 0 ? parsed : fallback;
}

function emit(payload) {
  console.log(JSON.stringify({
    event: "node-cicd-timing",
    schemaVersion: "v1",
    pipelineRun,
    taskRun,
    task,
    revision,
    source: "scripts/ci/runtime-ready.mjs",
    at: new Date().toISOString(),
    elapsedMs: Date.now() - startedAt,
    remainingMs: Math.max(0, deadline - Date.now()),
    ...payload
  }));
}

function sleep(ms) {
  return new Promise((resolve) => setTimeout(resolve, ms));
}

function safeJson(text) {
  try {
    return JSON.parse(text);
  } catch {
    return null;
  }
}

if (!host) {
  emit({ stage: "argo-refresh", status: "failed", reason: "kubernetes-service-host-missing", durationMs: 0 });
  process.exit(1);
}

const token = readFileSync("/var/run/secrets/kubernetes.io/serviceaccount/token", "utf8");
const ca = readFileSync("/var/run/secrets/kubernetes.io/serviceaccount/ca.crt");

function request(method, path, body, contentType = "application/json") {
  const payload = body ? JSON.stringify(body) : "";
  const headers = { Authorization: "Bearer " + token };
  if (payload) {
    headers["Content-Type"] = contentType;
    headers["Content-Length"] = Buffer.byteLength(payload);
  }
  return new Promise((resolve, reject) => {
    const req = https.request({ host, port, method, path, ca, headers }, (res) => {
      let data = "";
      res.setEncoding("utf8");
      res.on("data", (chunk) => { data += chunk; });
      res.on("end", () => resolve({ statusCode: res.statusCode || 0, body: data, json: safeJson(data) }));
    });
    req.on("error", reject);
    if (payload) req.write(payload);
    req.end();
  });
}

function listPath(group, version, namespace, resource) {
  if (group) return "/apis/" + group + "/" + version + "/namespaces/" + encodeURIComponent(namespace) + "/" + resource;
  return "/api/" + version + "/namespaces/" + encodeURIComponent(namespace) + "/" + resource;
}

async function runtimeItems() {
  const [deployments, statefulsets] = await Promise.all([
    request("GET", listPath("apps", "v1", runtimeNamespace, "deployments")),
    request("GET", listPath("apps", "v1", runtimeNamespace, "statefulsets"))
  ]);
  if (deployments.statusCode === 403 || statefulsets.statusCode === 403) return { status: "rbac-denied", items: [] };
  const items = [...(deployments.json?.items || []), ...(statefulsets.json?.items || [])]
    .filter((item) => item.metadata?.labels?.["hwlab.pikastech.local/gitops-target"] === gitopsTarget);
  return { status: "ok", items };
}

function runtimeSourceCommit(item) {
  return item.spec?.template?.metadata?.labels?.["hwlab.pikastech.local/gitops-render-source-commit"] ||
    item.spec?.template?.metadata?.annotations?.["hwlab.pikastech.local/gitops-render-source-commit"] ||
    item.metadata?.labels?.["hwlab.pikastech.local/gitops-render-source-commit"] ||
    item.metadata?.annotations?.["hwlab.pikastech.local/gitops-render-source-commit"] ||
    item.spec?.template?.metadata?.labels?.["hwlab.pikastech.local/source-commit"] ||
    item.spec?.template?.metadata?.annotations?.["hwlab.pikastech.local/source-commit"] ||
    item.metadata?.labels?.["hwlab.pikastech.local/source-commit"] ||
    item.metadata?.annotations?.["hwlab.pikastech.local/source-commit"] ||
    null;
}

function serviceIdForRuntimeItem(item) {
  return item.metadata?.labels?.["hwlab.pikastech.local/service-id"] ||
    item.spec?.template?.metadata?.labels?.["hwlab.pikastech.local/service-id"] ||
    item.metadata?.labels?.["app.kubernetes.io/name"] ||
    item.metadata?.name ||
    null;
}

function readObservedServiceIds() {
  try {
    const plan = JSON.parse(readFileSync("/workspace/source/affected-services.json", "utf8"));
    const rolloutServices = Array.isArray(plan.rolloutServices) ? plan.rolloutServices.filter(Boolean) : [];
    if (rolloutServices.length === 0) return null;
    return new Set(rolloutServices);
  } catch {
    return null;
  }
}

const observedServiceIds = readObservedServiceIds();
const readinessMode = observedServiceIds ? "service-rollout" : "infra-runtime-change";

function observedRuntimeItems(items) {
  if (!observedServiceIds) return items;
  return items.filter((item) => observedServiceIds.has(serviceIdForRuntimeItem(item)));
}

function readyWorkload(item) {
  const specReplicas = item.spec?.replicas ?? 1;
  if (specReplicas === 0) return { ready: true, skipped: true };
  const status = item.status || {};
  const desired = specReplicas;
  const ready = status.readyReplicas ?? 0;
  const updated = status.updatedReplicas ?? 0;
  const observedGeneration = status.observedGeneration ?? 0;
  const generation = item.metadata?.generation ?? 0;
  return { ready: observedGeneration >= generation && ready >= desired && updated >= desired, desired, readyReplicas: ready, updatedReplicas: updated };
}

function maybeEmitProgress(state, stage, stageStartedAt, payload) {
  const now = Date.now();
  if (now - state.lastProgressAt < progressEveryMs) return;
  state.lastProgressAt = now;
  emit({ stage, status: "progress", durationMs: now - stageStartedAt, readinessMode, ...payload });
}

async function waitForArgoRefresh() {
  const stageStartedAt = Date.now();
  const progressState = { lastProgressAt: 0 };
  while (Date.now() < deadline) {
    const runtime = await runtimeItems();
    if (runtime.status === "rbac-denied") return fail("argo-refresh", "runtime-observer-rbac-denied", stageStartedAt, { readinessMode });
    const observed = observedRuntimeItems(runtime.items);
    const pending = observed
      .map((item) => ({ kind: item.kind, name: item.metadata?.name, sourceCommit: runtimeSourceCommit(item) }))
      .filter((item) => item.sourceCommit !== revision);
    if (observed.length > 0 && pending.length === 0) {
      emit({ stage: "argo-refresh", status: "succeeded", durationMs: Date.now() - stageStartedAt, readinessMode, workloadCount: runtime.items.length, observedCount: observed.length });
      return true;
    }
    maybeEmitProgress(progressState, "argo-refresh", stageStartedAt, { workloadCount: runtime.items.length, observedCount: observed.length, pending: pending.slice(0, 8) });
    await sleep(Math.min(pollMs, Math.max(0, deadline - Date.now())));
  }
  return fail("argo-refresh", "source-commit-refresh-timeout", stageStartedAt, { readinessMode }, "timeout");
}

async function waitForWorkloads() {
  const stageStartedAt = Date.now();
  const progressState = { lastProgressAt: 0 };
  while (Date.now() < deadline) {
    const runtime = await runtimeItems();
    if (runtime.status === "rbac-denied") return fail("workload-ready", "runtime-observer-rbac-denied", stageStartedAt, { readinessMode });
    const observed = observedRuntimeItems(runtime.items);
    const blocked = observed.map((item) => ({ kind: item.kind, name: item.metadata?.name, ...readyWorkload(item) })).filter((item) => !item.ready);
    if (observed.length > 0 && blocked.length === 0) {
      emit({ stage: "workload-ready", status: "succeeded", durationMs: Date.now() - stageStartedAt, readinessMode, workloadCount: runtime.items.length, observedCount: observed.length });
      return true;
    }
    maybeEmitProgress(progressState, "workload-ready", stageStartedAt, { workloadCount: runtime.items.length, observedCount: observed.length, blocked: blocked.slice(0, 8) });
    await sleep(Math.min(pollMs, Math.max(0, deadline - Date.now())));
  }
  return fail("workload-ready", "runtime-ready-wall-clock-timeout", stageStartedAt, { readinessMode }, "timeout");
}

function fail(stage, reason, stageStartedAt, extra = {}, status = "failed") {
  emit({ stage, status, reason, durationMs: Date.now() - stageStartedAt, ...extra });
  process.exitCode = 1;
  return false;
}

emit({ stage: "runtime-ready", status: "started", readinessMode, observedServiceCount: observedServiceIds ? observedServiceIds.size : 0, timeoutMs });
try {
  if (await waitForArgoRefresh()) {
    await waitForWorkloads();
  }
} catch (error) {
  emit({ stage: "runtime-ready", status: "failed", reason: error instanceof Error ? error.message : String(error), durationMs: Date.now() - startedAt });
  process.exitCode = 1;
}