fix: don't block startup on subscribe (#359)

* fix: don't block startup on subscribe

* chore: comment

src/services/relay_renewal_job/refresh_topic_subscriptions.rs

@@ -7,7 +7,7 @@ use {
     },
     futures_util::{StreamExt, TryFutureExt, TryStreamExt},
     relay_client::http::Client,
-    relay_rpc::{domain::Topic, rpc::MAX_SUBSCRIPTION_BATCH_SIZE},
+    relay_rpc::domain::Topic,
     sqlx::PgPool,
     std::{sync::Arc, time::Instant},
     tokio::sync::Mutex,
@@ -46,36 +46,13 @@ pub async fn run(
     let topics_count = topics.len();
     info!("topics_count: {topics_count}");
 
-    // Collect each batch into its own vec, since `batch_subscribe` would convert
-    // them anyway.
-    let topic_batches = topics
-        .chunks(MAX_SUBSCRIPTION_BATCH_SIZE)
-        .map(|chunk| chunk.to_vec())
-        .collect::<Vec<_>>();
-
     // Limit concurrency to avoid overwhelming the relay with requests.
-    const REQUEST_CONCURRENCY: usize = 200;
-
-    futures_util::stream::iter(topic_batches)
-        .map(|topic_batch| {
-            // Map result to an unsized type to avoid allocation when collecting,
-            // as we don't care about subscription IDs.
-            client.batch_subscribe_blocking(topic_batch).map_ok(|_| ())
-        })
-        .buffer_unordered(REQUEST_CONCURRENCY)
-        .try_collect::<Vec<_>>()
-        .await?;
-
-    let elapsed: u64 = start
-        .elapsed()
-        .as_millis()
-        .try_into()
-        .expect("No error getting ms of elapsed time");
-    info!("resubscribe took {elapsed}ms");
+    const REQUEST_CONCURRENCY: usize = 25;
 
     // If operation already running, don't start another one
     let mut operation_running = renew_all_topics_lock.lock().await;
     if !*operation_running {
+        info!("Starting renew operation");
         *operation_running = true;
         // Renew all subscription TTLs.
         // This can take a long time (e.g. 2 hours), so cannot block server startup.
@@ -86,7 +63,9 @@ pub async fn run(
             async move {
                 let client = &client;
                 let metrics = metrics.as_ref();
-                let start = Instant::now();
+
+                // Using `batch_subscription` was removed in https://github.com/WalletConnect/notify-server/pull/359
+                // We can't really use this right now because we are also extending the topic TTL which could take longer than the 5m TTL
                 let result = futures_util::stream::iter(topics)
                     .map(|topic| async move {
                         // Subscribe a second time as the initial subscription above may have expired
@@ -101,7 +80,7 @@ pub async fn run(
                     })
                     // Above we want to resubscribe as quickly as possible so use a high concurrency value
                     // But here we prefer stability and are OK with a lower value
-                    .buffer_unordered(25)
+                    .buffer_unordered(REQUEST_CONCURRENCY)
                     .try_collect::<Vec<_>>()
                     .await;
                 let elapsed: u64 = start.elapsed().as_millis().try_into().unwrap();
@@ -114,19 +93,25 @@ pub async fn run(
                     info!("Success renewing all topic subscriptions in {elapsed}ms");
                 }
                 *renew_all_topics_lock.lock().await = false;
+
+                if let Some(metrics) = metrics {
+                    let ctx = Context::current();
+                    metrics.subscribed_project_topics.observe(
+                        &ctx,
+                        project_topics_count as u64,
+                        &[],
+                    );
+                    metrics.subscribed_subscriber_topics.observe(
+                        &ctx,
+                        subscriber_topics_count as u64,
+                        &[],
+                    );
+                    metrics.subscribe_latency.record(&ctx, elapsed, &[]);
+                }
             }
         });
-    }
-
-    if let Some(metrics) = metrics {
-        let ctx = Context::current();
-        metrics
-            .subscribed_project_topics
-            .observe(&ctx, project_topics_count as u64, &[]);
-        metrics
-            .subscribed_subscriber_topics
-            .observe(&ctx, subscriber_topics_count as u64, &[]);
-        metrics.subscribe_latency.record(&ctx, elapsed, &[]);
+    } else {
+        info!("Renew operation already running");
     }
 
     Ok(())