trac-msb 0.2.24 → 0.2.27

package/src/index.js

@@ -29,6 +29,8 @@ import {safeDecodeApplyOperation, safeEncodeApplyOperation} from "./utils/protob
 import PartialTransactionValidator from "./core/network/protocols/shared/validators/PartialTransactionValidator.js";
 import PartialTransferValidator from "./core/network/protocols/shared/validators/PartialTransferValidator.js";
 import { BroadcastError, ValidationError } from "./utils/errors.js";
+import { ReplicationStallGuard } from "./core/network/services/ReplicationStallGuard.js";
+import { hypercoreStallfixLine, hypercoreStallfixStatus } from "./utils/hypercoreStallfix.js";
 
 export class MainSettlementBus extends ReadyResource {
     #store;
@@ -37,6 +39,7 @@ export class MainSettlementBus extends ReadyResource {
     #state;
     #isClosing = false;
     #config
+    #replicationGuard = null;
 
     /**
      * @param {import("./config/config.js").Config} config
@@ -87,6 +90,13 @@ export class MainSettlementBus extends ReadyResource {
             this.#wallet,
         );
 
+        console.log(hypercoreStallfixLine('MSB', this.#state.base.core));
+        this.#replicationGuard = new ReplicationStallGuard(this.#state.base, () => this.#network?.swarm ?? null, {
+            timeout: this.#config.replicationStallTimeout,
+            refresh: () => this.#network?.swarm?.status?.(this.#config.channel)?.refresh?.()
+        });
+        this.#replicationGuard.start();
+
         const adminEntry = await this.#state.getAdminEntry();
         await this.#setUpRoleAutomatically(adminEntry);
 
@@ -106,6 +116,7 @@ export class MainSettlementBus extends ReadyResource {
         console.log("Closing everything gracefully... This may take a moment.");
 
         this.#isClosing = true;
+        this.#replicationGuard?.stop();
         await this.#network.close();
 
         await sleep(100);
@@ -121,6 +132,18 @@ export class MainSettlementBus extends ReadyResource {
         await sleep(100);
     }
 
+    /**
+     * Replication health: whether the loaded Hypercore carries the stall fix, and the stall guard state
+     * ('ok' | 'behind' | 'stalled'). Read-only.
+     */
+    replicationStatus() {
+        return {
+            hypercoreStallfix: hypercoreStallfixStatus(this.#state?.base?.core, this.#store),
+            stall: this.#replicationGuard?.status() ?? null,
+            connections: this.#network?.swarm?.connections?.size ?? 0
+        };
+    }
+
     async destroy() {
         return await this.close();
     }
@@ -316,6 +339,11 @@ export class MainSettlementBus extends ReadyResource {
             console.log("---------- flags ----------");
             console.log(`isIndexer: ${this.#state?.isIndexer?.() ?? "unset"}`);
             console.log(`isWriter: ${this.#state?.isWritable?.() ?? "unset"}`);
+            const replication = this.replicationStatus();
+            console.log("---------- replication ----------");
+            console.log(`hypercoreStallfix: ${replication.hypercoreStallfix.patched ? replication.hypercoreStallfix.marker : "NOT ACTIVE"}` +
+                ` (expired requests: ${replication.hypercoreStallfix.expiredRequests ?? "n/a"}, timed pauses: ${replication.hypercoreStallfix.invalidOperationPauses ?? "n/a"})`);
+            console.log(`replicationStall: ${replication.stall?.state ?? "off"} (reconnects: ${replication.stall?.reconnects ?? 0})`);
         } catch (error) {
             console.error("Error during DAG monitoring:", error.message);
         }

package/src/utils/hypercoreStallfix.js

@@ -0,0 +1,53 @@
+// Runtime view of the install-time Hypercore patch (scripts/hypercore-stallfix/). It looks at the Replicator class
+// of a live core, so it reports the Hypercore copy that is actually loaded, not the one npm resolution suggests.
+export const HYPERCORE_STALLFIX_MARKER = 'openmayhem-stallfix-v2';
+
+/**
+ * @param {object} core a Hypercore session (for MSB: state.base.core, the system view)
+ * @param {object|null} store optional Corestore; when given, request counters are summed over its open cores
+ */
+export function hypercoreStallfixStatus(core, store = null) {
+    const Replicator = core?.core?.replicator?.constructor ?? null;
+    const marker = typeof Replicator?.STALLFIX === 'string' ? Replicator.STALLFIX : null;
+    const status = {
+        patched: marker === HYPERCORE_STALLFIX_MARKER,
+        marker,
+        expiredRequests: null,
+        notAvailableSent: null,
+        notAvailableSuppressed: null,
+        invalidOperationPauses: null,
+        invalidOperationPermanent: null
+    };
+    if (marker === null) return status;
+
+    let expired = 0;
+    let sent = 0;
+    let suppressed = 0;
+    let opPauses = 0;
+    let opPermanent = 0;
+    const cores = store?.cores?.map;
+    const list = cores && typeof cores.values === 'function' ? cores.values() : [core.core];
+    for (const c of list) {
+        const stats = c?.replicator?._sfStats;
+        if (!stats) continue;
+        expired += stats.expired;
+        sent += stats.notAvailableSent;
+        suppressed += stats.notAvailableSuppressed;
+        opPauses += stats.invalidOperationPauses ?? 0; // 11.18.3 only (11.8.3 never pauses on INVALID_OPERATION)
+        opPermanent += stats.invalidOperationPermanent ?? 0;
+    }
+    status.expiredRequests = expired;
+    status.notAvailableSent = sent;
+    status.notAvailableSuppressed = suppressed;
+    status.invalidOperationPauses = opPauses;
+    status.invalidOperationPermanent = opPermanent;
+    return status;
+}
+
+/** One startup line for logs/alerts. */
+export function hypercoreStallfixLine(label, core) {
+    const { patched, marker } = hypercoreStallfixStatus(core);
+    if (patched) return `[${label}] Hypercore replication stall fix: active (${marker}); closed-stream guard #911 active`;
+    return `[${label}] WARNING: Hypercore replication stall fix NOT active in the loaded hypercore` +
+        ` (marker: ${marker ?? 'none'}). Reinstall without --ignore-scripts or run scripts/hypercore-stallfix/apply.mjs.`;
+}