Merge branch 'main' into dev

server/private/cleanup.ts

@@ -15,8 +15,10 @@ import { rateLimitService } from "#private/lib/rateLimit";
 import { cleanup as wsCleanup } from "#private/routers/ws";
 import { flushBandwidthToDb } from "@server/routers/newt/handleReceiveBandwidthMessage";
 import { flushSiteBandwidthToDb } from "@server/routers/gerbil/receiveBandwidth";
+import { stopPingAccumulator } from "@server/routers/newt/pingAccumulator";
 
 async function cleanup() {
+    await stopPingAccumulator();
     await flushBandwidthToDb();
     await flushSiteBandwidthToDb();
     await rateLimitService.cleanup();

server/private/lib/cache.ts

@@ -1,3 +1,16 @@
+/*
+ * This file is part of a proprietary work.
+ *
+ * Copyright (c) 2025 Fossorial, Inc.
+ * All rights reserved.
+ *
+ * This file is licensed under the Fossorial Commercial License.
+ * You may not use this file except in compliance with the License.
+ * Unauthorized use, copying, modification, or distribution is strictly prohibited.
+ *
+ * This file is not licensed under the AGPLv3.
+ */
+
 import NodeCache from "node-cache";
 import logger from "@server/logger";
 import { redisManager } from "@server/private/lib/redis";

server/private/lib/tokenCache.ts

@@ -0,0 +1,77 @@
+/*
+ * This file is part of a proprietary work.
+ *
+ * Copyright (c) 2025 Fossorial, Inc.
+ * All rights reserved.
+ *
+ * This file is licensed under the Fossorial Commercial License.
+ * You may not use this file except in compliance with the License.
+ * Unauthorized use, copying, modification, or distribution is strictly prohibited.
+ *
+ * This file is not licensed under the AGPLv3.
+ */
+
+import redisManager from "#private/lib/redis";
+import { encrypt, decrypt } from "@server/lib/crypto";
+import logger from "@server/logger";
+
+/**
+ * Returns a cached plaintext token from Redis if one exists and decrypts
+ * cleanly, otherwise calls `createSession` to mint a fresh token, stores the
+ * encrypted value in Redis with the given TTL, and returns it.
+ *
+ * Failures at the Redis layer are non-fatal – the function always falls
+ * through to session creation so the caller is never blocked by a Redis outage.
+ *
+ * @param cacheKey   Unique Redis key, e.g. `"newt:token_cache:abc123"`
+ * @param secret     Server secret used for AES encryption/decryption
+ * @param ttlSeconds Cache TTL in seconds (should match session expiry)
+ * @param createSession Factory that mints a new session and returns its raw token
+ */
+export async function getOrCreateCachedToken(
+    cacheKey: string,
+    secret: string,
+    ttlSeconds: number,
+    createSession: () => Promise<string>
+): Promise<string> {
+    if (redisManager.isRedisEnabled()) {
+        try {
+            const cached = await redisManager.get(cacheKey);
+            if (cached) {
+                const token = decrypt(cached, secret);
+                if (token) {
+                    logger.debug(`Token cache hit for key: ${cacheKey}`);
+                    return token;
+                }
+                // Decryption produced an empty string – treat as a miss
+                logger.warn(
+                    `Token cache decryption returned empty string for key: ${cacheKey}, treating as miss`
+                );
+            }
+        } catch (e) {
+            logger.warn(
+                `Token cache read/decrypt failed for key ${cacheKey}, falling through to session creation:`,
+                e
+            );
+        }
+    }
+
+    const token = await createSession();
+
+    if (redisManager.isRedisEnabled()) {
+        try {
+            const encrypted = encrypt(token, secret);
+            await redisManager.set(cacheKey, encrypted, ttlSeconds);
+            logger.debug(
+                `Token cached in Redis for key: ${cacheKey} (TTL ${ttlSeconds}s)`
+            );
+        } catch (e) {
+            logger.warn(
+                `Token cache write failed for key ${cacheKey} (session was still created):`,
+                e
+            );
+        }
+    }
+
+    return token;
+}

server/private/routers/remoteExitNode/getRemoteExitNodeToken.ts

@@ -23,8 +23,10 @@ import { z } from "zod";
 import { fromError } from "zod-validation-error";
 import {
     createRemoteExitNodeSession,
-    validateRemoteExitNodeSessionToken
+    validateRemoteExitNodeSessionToken,
+    EXPIRES
 } from "#private/auth/sessions/remoteExitNode";
+import { getOrCreateCachedToken } from "@server/private/lib/tokenCache";
 import { verifyPassword } from "@server/auth/password";
 import logger from "@server/logger";
 import config from "@server/lib/config";
@@ -103,14 +105,23 @@ export async function getRemoteExitNodeToken(
             );
         }
 
-        const resToken = generateSessionToken();
-        await createRemoteExitNodeSession(
-            resToken,
-            existingRemoteExitNode.remoteExitNodeId
+        // Return a cached token if one exists to prevent thundering herd on
+        // simultaneous restarts; falls back to creating a fresh session when
+        // Redis is unavailable or the cache has expired.
+        const resToken = await getOrCreateCachedToken(
+            `remote_exit_node:token_cache:${existingRemoteExitNode.remoteExitNodeId}`,
+            config.getRawConfig().server.secret!,
+            Math.floor(EXPIRES / 1000),
+            async () => {
+                const token = generateSessionToken();
+                await createRemoteExitNodeSession(
+                    token,
+                    existingRemoteExitNode.remoteExitNodeId
+                );
+                return token;
+            }
         );
 
-        // logger.debug(`Created RemoteExitNode token response: ${JSON.stringify(resToken)}`);
-
         return response<{ token: string }>(res, {
             data: {
                 token: resToken

server/private/routers/ws/ws.ts

@@ -19,17 +19,14 @@ import { Socket } from "net";
 import {
     Newt,
     newts,
-    NewtSession,
-    olms,
     Olm,
-    OlmSession,
+    olms,
     RemoteExitNode,
-    RemoteExitNodeSession,
     remoteExitNodes,
-    sites
 } from "@server/db";
 import { eq } from "drizzle-orm";
 import { db } from "@server/db";
+import { recordPing } from "@server/routers/newt/pingAccumulator";
 import { validateNewtSessionToken } from "@server/auth/sessions/newt";
 import { validateOlmSessionToken } from "@server/auth/sessions/olm";
 import logger from "@server/logger";
@@ -197,11 +194,7 @@ const connectedClients: Map<string, AuthenticatedWebSocket[]> = new Map();
 // Config version tracking map (local to this node, resets on server restart)
 const clientConfigVersions: Map<string, number> = new Map();
 
-// Tracks the last Unix timestamp (seconds) at which a ping was flushed to the
-// DB for a given siteId. Resets on server restart which is fine – the first
-// ping after startup will always write, re-establishing the online state.
-const lastPingDbWrite: Map<number, number> = new Map();
-const PING_DB_WRITE_INTERVAL = 45; // seconds
+
 
 // Recovery tracking
 let isRedisRecoveryInProgress = false;
@@ -853,32 +846,16 @@ const setupConnection = async (
         );
     });
 
-    // Handle WebSocket protocol-level pings from older newt clients that do
-    // not send application-level "newt/ping" messages. Update the site's
-    // online state and lastPing timestamp so the offline checker treats them
-    // the same as modern newt clients.
     if (clientType === "newt") {
         const newtClient = client as Newt;
-        ws.on("ping", async () => {
+        ws.on("ping", () => {
             if (!newtClient.siteId) return;
-            const now = Math.floor(Date.now() / 1000);
-            const lastWrite = lastPingDbWrite.get(newtClient.siteId) ?? 0;
-            if (now - lastWrite < PING_DB_WRITE_INTERVAL) return;
-            lastPingDbWrite.set(newtClient.siteId, now);
-            try {
-                await db
-                    .update(sites)
-                    .set({
-                        online: true,
-                        lastPing: now
-                    })
-                    .where(eq(sites.siteId, newtClient.siteId));
-            } catch (error) {
-                logger.error(
-                    "Error updating newt site online state on WS ping",
-                    { error }
-                );
-            }
+            // Record the ping in the accumulator instead of writing to the
+            // database on every WS ping frame. The accumulator flushes all
+            // pending pings in a single batched UPDATE every ~10s, which
+            // prevents connection pool exhaustion under load (especially
+            // with cross-region latency to the database).
+            recordPing(newtClient.siteId);
         });
     }