@ts-cloud/core 0.16.14 → 0.16.15

package/dist/types.d.ts

@@ -977,6 +977,39 @@ export interface SharedPathSpec {
 }
 /** A shared path: a release-relative path, or {@link SharedPathSpec}. */
 export type SharedPathEntry = string | SharedPathSpec;
+/**
+ * The recurring liveness probe for a ported server-app site: a systemd timer
+ * that asks the service for an HTTP response on `127.0.0.1:<port>` and
+ * restarts it when it has stopped answering. Every field is optional.
+ */
+export interface SiteLivenessConfig {
+    /** Path to request. @default the health check path, else '/' */
+    path?: string;
+    /** Seconds between checks. @default 60 */
+    intervalSeconds?: number;
+    /** Consecutive failed checks, after the startup grace, before a restart. @default 3 */
+    failuresBeforeRestart?: number;
+    /**
+     * Seconds one check waits for an HTTP response. A slow answer inside this
+     * budget is an answer, and counts as alive. @default 10
+     */
+    timeoutSeconds?: number;
+    /**
+     * Seconds after a unit (re)starts during which a failed check is not held
+     * against it - the equivalent of a Kubernetes startupProbe. Must be longer
+     * than the slowest honest boot: a server that binds its port and then warms
+     * up for minutes is starting, not wedged. `0` disables the grace.
+     * @default 600
+     */
+    startupGraceSeconds?: number;
+    /**
+     * Ceiling on the back-off between consecutive liveness restarts. The second
+     * restart in a row waits at least 5 minutes after the first, and each one
+     * after that doubles, up to this. `0` disables the back-off.
+     * @default 3600
+     */
+    maxBackoffSeconds?: number;
+}
 export interface SiteConfig {
     /**
      * Directory to deploy.
@@ -1184,6 +1217,19 @@ export interface SiteConfig {
      * exactly that. Leave unset for anything that stops immediately.
      */
     stopTimeout?: string;
+    /**
+     * Recurring liveness probe for a ported server-app site. `Restart=always`
+     * only covers a process that exits; this covers one that is alive and no
+     * longer answering. A unit is never restarted inside its startup grace, and
+     * consecutive restarts back off, so a slow boot cannot turn into a restart
+     * loop.
+     *
+     * Set `false` to opt out, for a service that answers nothing on its port or
+     * one where a restart is more dangerous than an outage.
+     * @default enabled: `healthCheck.path` or '/', every 60s, 10s timeout,
+     * restart after 3 consecutive failures once the unit is 10 minutes old
+     */
+    liveness?: false | SiteLivenessConfig;
     /**
      * SSR only. tar `--exclude` patterns applied when packaging the release
      * tarball. Keep host-specific / heavy paths out of the artifact — most

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@ts-cloud/core",
   "type": "module",
-  "version": "0.16.14",
+  "version": "0.16.15",
   "description": "Core CloudFormation generation library for ts-cloud",
   "author": "Chris Breuer <chris@stacksjs.com>",
   "license": "MIT",
@@ -31,7 +31,7 @@
     "typecheck": "tsc --noEmit"
   },
   "dependencies": {
-    "@ts-cloud/aws-types": "0.16.14"
+    "@ts-cloud/aws-types": "0.16.15"
   },
   "devDependencies": {
     "typescript": "^7.0.2"