Merge branch 'master' into prevent-duplicate-lightning

Reviewed-on: coracle/caravel#49
Co-authored-by: userAdityaa <aditya.chaudhary1558@gmail.com>
Co-committed-by: userAdityaa <aditya.chaudhary1558@gmail.com>

backend/migrations/0003_invoice_nwc_payment_guard.sql

@@ -0,0 +1,11 @@
+CREATE TABLE IF NOT EXISTS invoice_nwc_payment (
+  invoice_id TEXT PRIMARY KEY,
+  tenant_pubkey TEXT NOT NULL,
+  state TEXT NOT NULL CHECK (state IN ('pending', 'paid')),
+  created_at INTEGER NOT NULL,
+  updated_at INTEGER NOT NULL,
+  FOREIGN KEY (tenant_pubkey) REFERENCES tenant(pubkey)
+);
+
+CREATE INDEX IF NOT EXISTS idx_invoice_nwc_payment_tenant_pubkey
+ON invoice_nwc_payment (tenant_pubkey);

backend/spec/infra.md

@@ -15,12 +15,15 @@ Members:
 ## `pub async fn start(self)`
 
 - Subscribes to `command.notify`
+- On startup, schedules delayed sync retries for relays whose `sync_error` is non-empty.
 - Loops on `rx.recv()`, calling `handle_activity` for each received `Activity`.
 
 ## `async fn handle_activity(&self, activity: &Activity)`
 
-- For `create_relay`, `update_relay`, `activate_relay`, or `deactivate_relay` activity, calls `sync_and_report`.
-- All other activity types are ignored (e.g. `fail_relay_sync`, `complete_relay_sync`).
+- For `create_relay`, `update_relay`, `activate_relay`, or `deactivate_relay` activity, calls `sync_and_report` immediately.
+- For `fail_relay_sync`, schedules a delayed retry using exponential backoff based on consecutive failures for the relay.
+- Retry scheduling stops after the configured max attempts to avoid infinite retry loops.
+- Other activity types are ignored (e.g. `complete_relay_sync`).
 
 ## `async fn sync_and_report(&self, relay: &Relay, is_new: bool)`
 

backend/src/billing.rs

@@ -17,6 +17,9 @@ type HmacSha256 = Hmac<Sha256>;
 const STRIPE_API: &str = "https://api.stripe.com/v1";
 const COINBASE_SPOT_API: &str = "https://api.coinbase.com/v2/prices";
 const WEBHOOK_TOLERANCE_SECS: i64 = 300;
+const MANUAL_LIGHTNING_PAYMENT_DM: &str = "Payment is due for your relay subscription. Please visit the application to complete a manual Lightning payment.";
+const NWC_ERROR_DM_PREFIX: &str = "NWC auto-payment failed:";
+const NWC_ERROR_DM_MAX_CHARS: usize = 240;
 
 #[derive(Debug)]
 pub enum InvoiceLookupError {
@@ -75,12 +78,17 @@ struct CoinbaseSpotPriceData {
     amount: String,
 }
 
+enum NwcInvoicePaymentOutcome {
+    Paid,
+    Fallback(anyhow::Error),
+    Pending(anyhow::Error),
+}
+
 #[derive(Clone)]
 pub struct Billing {
     nwc_url: String,
     stripe_secret_key: String,
     stripe_webhook_secret: String,
-    btc_quote_api_base: String,
     http: reqwest::Client,
     query: Query,
     command: Command,
@@ -98,13 +106,10 @@ impl Billing {
         if stripe_webhook_secret.trim().is_empty() {
             panic!("missing STRIPE_WEBHOOK_SECRET environment variable");
         }
-        let btc_quote_api_base =
-            std::env::var("BTC_PRICE_API_BASE").unwrap_or_else(|_| COINBASE_SPOT_API.to_string());
         Self {
             nwc_url,
             stripe_secret_key,
             stripe_webhook_secret,
-            btc_quote_api_base,
             http: reqwest::Client::new(),
             query,
             command,
@@ -261,6 +266,23 @@ impl Billing {
         Ok(())
     }
 
+    async fn existing_invoice_nwc_payment_outcome(
+        &self,
+        invoice_id: &str,
+    ) -> Result<Option<NwcInvoicePaymentOutcome>> {
+        let state = self.query.get_invoice_nwc_payment_state(invoice_id).await?;
+        match state.as_deref() {
+            Some("paid") => Ok(Some(NwcInvoicePaymentOutcome::Paid)),
+            Some("pending") => Ok(Some(NwcInvoicePaymentOutcome::Pending(anyhow!(
+                "invoice {invoice_id} has a pending NWC reconciliation; refusing to create a new Lightning charge"
+            )))),
+            Some(other) => Err(anyhow!(
+                "unknown invoice_nwc_payment state '{other}' for invoice {invoice_id}"
+            )),
+            None => Ok(None),
+        }
+    }
+
     pub async fn handle_webhook(&self, payload: &str, signature: &str) -> Result<()> {
         self.verify_webhook_signature(payload, signature)?;
 
@@ -360,24 +382,54 @@ impl Billing {
             return Ok(());
         };
 
+        let mut nwc_error_for_dm: Option<String> = None;
+
         // 1. NWC auto-pay: if the tenant has a nwc_url
         if !tenant.nwc_url.is_empty() {
             match self
-                .nwc_pay_invoice(amount_due, currency, &tenant.nwc_url)
-                .await
+                .nwc_pay_invoice(
+                    invoice_id,
+                    &tenant.pubkey,
+                    amount_due,
+                    currency,
+                    &tenant.nwc_url,
+                )
+                .await?
             {
-                Ok(()) => {
-                    self.stripe_pay_invoice_out_of_band(invoice_id).await?;
-                    self.command.clear_tenant_nwc_error(&tenant.pubkey).await?;
+                NwcInvoicePaymentOutcome::Paid => {
+                    self.mark_invoice_paid_out_of_band_after_nwc(invoice_id, &tenant.pubkey)
+                        .await?;
                     return Ok(());
                 }
-                Err(e) => {
+                NwcInvoicePaymentOutcome::Fallback(e) => {
                     let error_msg = format!("{e}");
                     self.command
                         .set_tenant_nwc_error(&tenant.pubkey, &error_msg)
                         .await?;
+                    tracing::warn!(
+                        error = %e,
+                        tenant_pubkey = %tenant.pubkey,
+                        stripe_customer_id,
+                        invoice_id,
+                        "nwc auto-payment failed for invoice.created"
+                    );
+                    nwc_error_for_dm = summarize_nwc_error_for_dm(&error_msg);
                     // Fall through to next option
                 }
+                NwcInvoicePaymentOutcome::Pending(e) => {
+                    let error_msg = format!("{e}");
+                    self.command
+                        .set_tenant_nwc_error(&tenant.pubkey, &error_msg)
+                        .await?;
+                    tracing::error!(
+                        error = %e,
+                        tenant_pubkey = %tenant.pubkey,
+                        stripe_customer_id,
+                        invoice_id,
+                        "nwc auto-payment requires reconciliation before retry"
+                    );
+                    return Err(e);
+                }
             }
         }
 
@@ -390,12 +442,8 @@ impl Billing {
         }
 
         // 3. Manual payment: send a DM
-        self.robot
-            .send_dm(
-                &tenant.pubkey,
-                "Payment is due for your relay subscription. Please visit the application to complete a manual Lightning payment.",
-            )
-            .await?;
+        let dm_message = manual_lightning_payment_dm(nwc_error_for_dm.as_deref());
+        self.robot.send_dm(&tenant.pubkey, &dm_message).await?;
 
         Ok(())
     }
@@ -638,11 +686,13 @@ impl Billing {
     pub async fn stripe_create_customer(&self, tenant_pubkey: &str) -> Result<String> {
         let short_pubkey: String = tenant_pubkey.chars().take(12).collect();
         let display_name = format!("Caravel tenant {short_pubkey}");
+        let idempotency_key = self.idempotency_key(&["create_customer", tenant_pubkey]);
 
         let resp = self
             .http
             .post(format!("{STRIPE_API}/customers"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .form(&[
                 ("name", display_name.as_str()),
                 ("metadata[tenant_pubkey]", tenant_pubkey),
@@ -743,21 +793,28 @@ impl Billing {
             }
 
             match self
-                .nwc_pay_invoice(amount_due, currency, &tenant.nwc_url)
-                .await
+                .nwc_pay_invoice(
+                    invoice_id,
+                    &tenant.pubkey,
+                    amount_due,
+                    currency,
+                    &tenant.nwc_url,
+                )
+                .await?
             {
-                Ok(()) => {
-                    if let Err(e) = self.stripe_pay_invoice_out_of_band(invoice_id).await {
+                NwcInvoicePaymentOutcome::Paid => {
+                    if let Err(e) = self
+                        .mark_invoice_paid_out_of_band_after_nwc(invoice_id, &tenant.pubkey)
+                        .await
+                    {
                         tracing::error!(
                             error = %e,
                             invoice_id,
                             "failed to mark invoice paid out of band"
                         );
-                    } else {
-                        let _ = self.command.clear_tenant_nwc_error(&tenant.pubkey).await;
                     }
                 }
-                Err(e) => {
+                NwcInvoicePaymentOutcome::Fallback(e) => {
                     let error_msg = format!("{e}");
                     tracing::error!(
                         error = %e,
@@ -769,6 +826,18 @@ impl Billing {
                         .set_tenant_nwc_error(&tenant.pubkey, &error_msg)
                         .await;
                 }
+                NwcInvoicePaymentOutcome::Pending(e) => {
+                    let error_msg = format!("{e}");
+                    tracing::error!(
+                        error = %e,
+                        invoice_id,
+                        "outstanding invoice requires NWC reconciliation before retry"
+                    );
+                    let _ = self
+                        .command
+                        .set_tenant_nwc_error(&tenant.pubkey, &error_msg)
+                        .await;
+                }
             }
         }
 
@@ -829,15 +898,30 @@ impl Billing {
 
     // --- Stripe API helpers ---
 
+    fn idempotency_key(&self, parts: &[&str]) -> String {
+        let mut mac = HmacSha256::new_from_slice(self.stripe_secret_key.as_bytes())
+            .expect("HMAC accepts any key length");
+        for (i, part) in parts.iter().enumerate() {
+            if i > 0 {
+                mac.update(b":");
+            }
+            mac.update(part.as_bytes());
+        }
+        hex::encode(mac.finalize().into_bytes())
+    }
+
     async fn stripe_create_subscription(
         &self,
         customer_id: &str,
         price_id: &str,
     ) -> Result<(String, String)> {
+        let idempotency_key =
+            self.idempotency_key(&["create_subscription", customer_id, price_id]);
         let resp = self
             .http
             .post(format!("{STRIPE_API}/subscriptions"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .form(&[
                 ("customer", customer_id),
                 ("collection_method", "charge_automatically"),
@@ -864,10 +948,13 @@ impl Billing {
         subscription_id: &str,
         price_id: &str,
     ) -> Result<String> {
+        let idempotency_key =
+            self.idempotency_key(&["create_subscription_item", subscription_id, price_id]);
         let resp = self
             .http
             .post(format!("{STRIPE_API}/subscription_items"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .form(&[("subscription", subscription_id), ("price", price_id)])
             .send()
             .await?;
@@ -886,10 +973,13 @@ impl Billing {
         item_id: &str,
         price_id: &str,
     ) -> Result<String> {
+        let idempotency_key =
+            self.idempotency_key(&["update_subscription_item", item_id, price_id]);
         let resp = self
             .http
             .post(format!("{STRIPE_API}/subscription_items/{item_id}"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .form(&[("price", price_id)])
             .send()
             .await?;
@@ -926,9 +1016,11 @@ impl Billing {
     }
 
     async fn stripe_pay_invoice(&self, invoice_id: &str) -> Result<()> {
+        let idempotency_key = self.idempotency_key(&["pay_invoice", invoice_id]);
         self.http
             .post(format!("{STRIPE_API}/invoices/{invoice_id}/pay"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .send()
             .await?
             .error_for_status()?;
@@ -968,9 +1060,11 @@ impl Billing {
     }
 
     async fn stripe_pay_invoice_out_of_band(&self, invoice_id: &str) -> Result<()> {
+        let idempotency_key = self.idempotency_key(&["pay_invoice_oob", invoice_id]);
         self.http
             .post(format!("{STRIPE_API}/invoices/{invoice_id}/pay"))
             .bearer_auth(&self.stripe_secret_key)
+            .header("Idempotency-Key", idempotency_key)
             .form(&[("paid_out_of_band", "true")])
             .send()
             .await?
@@ -999,19 +1093,47 @@ impl Billing {
 
     // --- NWC helpers ---
 
+    async fn mark_invoice_paid_out_of_band_after_nwc(
+        &self,
+        invoice_id: &str,
+        tenant_pubkey: &str,
+    ) -> Result<()> {
+        self.stripe_pay_invoice_out_of_band(invoice_id).await?;
+        self.command.clear_tenant_nwc_error(tenant_pubkey).await?;
+        Ok(())
+    }
+
+    fn parse_nwc_uri(nwc_url: &str, role: &str) -> Result<NostrWalletConnectURI> {
+        nwc_url
+            .parse::<NostrWalletConnectURI>()
+            .map_err(|_| anyhow!("invalid {role} NWC URL"))
+    }
+
     async fn nwc_pay_invoice(
         &self,
+        invoice_id: &str,
+        tenant_pubkey: &str,
         amount_due_minor: i64,
         currency: &str,
         tenant_nwc_url: &str,
-    ) -> Result<()> {
-        let amount_msats = self.fiat_minor_to_msats(amount_due_minor, currency).await?;
+    ) -> Result<NwcInvoicePaymentOutcome> {
+        if let Some(existing_outcome) = self
+            .existing_invoice_nwc_payment_outcome(invoice_id)
+            .await?
+        {
+            return Ok(existing_outcome);
+        }
+
+        let amount_msats = match self.fiat_minor_to_msats(amount_due_minor, currency).await {
+            Ok(amount_msats) => amount_msats,
+            Err(error) => return Ok(NwcInvoicePaymentOutcome::Fallback(error)),
+        };
 
         // Create a bolt11 invoice using the system wallet (self.nwc_url)
-        let system_uri: NostrWalletConnectURI = self
-            .nwc_url
-            .parse()
-            .map_err(|_| anyhow!("invalid system NWC URL"))?;
+        let system_uri = match Self::parse_nwc_uri(&self.nwc_url, "system") {
+            Ok(system_uri) => system_uri,
+            Err(error) => return Ok(NwcInvoicePaymentOutcome::Fallback(error)),
+        };
         let system_nwc = NWC::new(system_uri);
 
         let make_req = MakeInvoiceRequest {
@@ -1021,29 +1143,61 @@ impl Billing {
             expiry: None,
         };
 
-        let invoice_response = system_nwc
-            .make_invoice(make_req)
-            .await
-            .map_err(|e| anyhow!("failed to create invoice: {e}"))?;
+        let invoice_response = system_nwc.make_invoice(make_req).await;
+
+        let invoice_response = match invoice_response {
+            Ok(invoice_response) => invoice_response,
+            Err(error) => {
+                system_nwc.shutdown().await;
+                return Ok(NwcInvoicePaymentOutcome::Fallback(anyhow!(
+                    "failed to create invoice: {error}"
+                )));
+            }
+        };
 
         system_nwc.shutdown().await;
 
         // Pay the bolt11 invoice using the tenant's wallet
-        let tenant_uri: NostrWalletConnectURI = tenant_nwc_url
-            .parse()
-            .map_err(|_| anyhow!("invalid tenant NWC URL"))?;
+        let tenant_uri = match Self::parse_nwc_uri(tenant_nwc_url, "tenant") {
+            Ok(tenant_uri) => tenant_uri,
+            Err(error) => return Ok(NwcInvoicePaymentOutcome::Fallback(error)),
+        };
+
+        if !self
+            .command
+            .insert_pending_invoice_nwc_payment(invoice_id, tenant_pubkey)
+            .await?
+        {
+            if let Some(existing_outcome) = self
+                .existing_invoice_nwc_payment_outcome(invoice_id)
+                .await?
+            {
+                return Ok(existing_outcome);
+            }
+            return Err(anyhow!(
+                "invoice_nwc_payment row missing after insert race for invoice {invoice_id}"
+            ));
+        }
+
         let tenant_nwc = NWC::new(tenant_uri);
 
         let pay_req = NwcPayInvoiceRequest::new(invoice_response.invoice);
 
-        tenant_nwc
-            .pay_invoice(pay_req)
-            .await
-            .map_err(|e| anyhow!("failed to pay invoice: {e}"))?;
+        let pay_result = tenant_nwc.pay_invoice(pay_req).await;
 
         tenant_nwc.shutdown().await;
 
-        Ok(())
+        match pay_result {
+            Ok(_) => match self.command.mark_invoice_nwc_payment_paid(invoice_id).await {
+                Ok(()) => Ok(NwcInvoicePaymentOutcome::Paid),
+                Err(error) => Ok(NwcInvoicePaymentOutcome::Pending(anyhow!(
+                    "invoice {invoice_id} was charged over NWC but failed to persist paid state: {error}"
+                ))),
+            },
+            Err(error) => Ok(NwcInvoicePaymentOutcome::Pending(anyhow!(
+                "invoice {invoice_id} NWC payment attempt requires reconciliation: {error}"
+            ))),
+        }
     }
 
     async fn fiat_minor_to_msats(&self, amount_due_minor: i64, currency: &str) -> Result<u64> {
@@ -1057,7 +1211,7 @@ impl Billing {
     }
 
     async fn fetch_btc_spot_price(&self, currency: &str) -> Result<f64> {
-        fetch_btc_spot_price_from_base(&self.http, &self.btc_quote_api_base, currency).await
+        fetch_btc_spot_price_from_base(&self.http, COINBASE_SPOT_API, currency).await
     }
 
     fn currency_minor_exponent(currency: &str) -> Result<u8> {
@@ -1101,6 +1255,31 @@ pub async fn fetch_btc_spot_price_from_base(
     Ok(amount)
 }
 
+fn summarize_nwc_error_for_dm(error: &str) -> Option<String> {
+    let normalized = error.split_whitespace().collect::<Vec<_>>().join(" ");
+    if normalized.is_empty() {
+        return None;
+    }
+
+    if normalized.chars().count() <= NWC_ERROR_DM_MAX_CHARS {
+        return Some(normalized);
+    }
+
+    let prefix_len = NWC_ERROR_DM_MAX_CHARS.saturating_sub(3);
+    let mut truncated = normalized.chars().take(prefix_len).collect::<String>();
+    truncated.push_str("...");
+    Some(truncated)
+}
+
+fn manual_lightning_payment_dm(nwc_error: Option<&str>) -> String {
+    match nwc_error {
+        Some(error) if !error.is_empty() => {
+            format!("{MANUAL_LIGHTNING_PAYMENT_DM}\n\n{NWC_ERROR_DM_PREFIX} {error}")
+        }
+        _ => MANUAL_LIGHTNING_PAYMENT_DM.to_string(),
+    }
+}
+
 pub fn fiat_minor_to_msats_from_quote(
     amount_due_minor: i64,
     currency: &str,
@@ -1398,4 +1577,5 @@ mod tests {
         assert_eq!(billing.stripe_secret_key, "sk_test_dummy");
         assert_eq!(billing.stripe_webhook_secret, "whsec_test_dummy");
     }
+
 }

backend/src/command.rs

@@ -313,6 +313,46 @@ impl Command {
         Ok(())
     }
 
+    pub async fn insert_pending_invoice_nwc_payment(
+        &self,
+        invoice_id: &str,
+        tenant_pubkey: &str,
+    ) -> Result<bool> {
+        let now = chrono::Utc::now().timestamp();
+        let result = sqlx::query(
+            "INSERT INTO invoice_nwc_payment (invoice_id, tenant_pubkey, state, created_at, updated_at)
+             VALUES (?, ?, 'pending', ?, ?)
+             ON CONFLICT(invoice_id) DO NOTHING",
+        )
+        .bind(invoice_id)
+        .bind(tenant_pubkey)
+        .bind(now)
+        .bind(now)
+        .execute(&self.pool)
+        .await?;
+
+        Ok(result.rows_affected() > 0)
+    }
+
+    pub async fn mark_invoice_nwc_payment_paid(&self, invoice_id: &str) -> Result<()> {
+        let now = chrono::Utc::now().timestamp();
+        let result = sqlx::query(
+            "UPDATE invoice_nwc_payment
+             SET state = 'paid', updated_at = ?
+             WHERE invoice_id = ?",
+        )
+        .bind(now)
+        .bind(invoice_id)
+        .execute(&self.pool)
+        .await?;
+
+        if result.rows_affected() == 0 {
+            anyhow::bail!("invoice_nwc_payment row missing for invoice_id: {invoice_id}");
+        }
+
+        Ok(())
+    }
+
     pub async fn set_tenant_past_due(&self, pubkey: &str) -> Result<()> {
         let now = chrono::Utc::now().timestamp();
         sqlx::query("UPDATE tenant SET past_due_at = ? WHERE pubkey = ?")

backend/src/infra.rs

@@ -1,10 +1,15 @@
 use anyhow::Result;
 use nostr_sdk::prelude::*;
+use std::time::Duration;
 
 use crate::command::Command;
 use crate::models::{Activity, RELAY_STATUS_DELINQUENT, RELAY_STATUS_INACTIVE, Relay};
 use crate::query::Query;
 
+const RELAY_SYNC_RETRY_BASE_DELAY_SECS: u64 = 30;
+const RELAY_SYNC_RETRY_MAX_DELAY_SECS: u64 = 15 * 60;
+const RELAY_SYNC_RETRY_MAX_ATTEMPTS: usize = 6;
+
 #[derive(Clone)]
 pub struct Infra {
     api_url: String,
@@ -48,6 +53,10 @@ impl Infra {
     pub async fn start(self) {
         let mut rx = self.command.notify.subscribe();
 
+        if let Err(e) = self.schedule_startup_retries().await {
+            tracing::error!(error = %e, "failed to schedule relay sync retries on startup");
+        }
+
         loop {
             match rx.recv().await {
                 Ok(activity) => {
@@ -66,15 +75,84 @@ impl Infra {
     async fn handle_activity(&self, activity: &Activity) -> Result<()> {
         let needs_sync = should_sync_relay_activity(activity.activity_type.as_str());
 
-        if needs_sync {
-            let Some(relay) = self.query.get_relay(&activity.resource_id).await? else {
-                return Ok(());
-            };
-
-            let is_new = relay.synced == 0;
-            self.sync_and_report(&relay, is_new).await;
+        if !needs_sync || activity.resource_type != "relay" {
+            return Ok(());
         }
 
+        if activity.activity_type == "fail_relay_sync" {
+            self.schedule_relay_sync_retry(&activity.resource_id, "activity")
+                .await?;
+            return Ok(());
+        }
+
+        let Some(relay) = self.query.get_relay(&activity.resource_id).await? else {
+            return Ok(());
+        };
+
+        let is_new = relay.synced == 0;
+        self.sync_and_report(&relay, is_new).await;
+
+        Ok(())
+    }
+
+    async fn schedule_startup_retries(&self) -> Result<()> {
+        let relays = self.query.list_relays_with_sync_error().await?;
+
+        for relay in relays {
+            self.schedule_relay_sync_retry(&relay.id, "startup").await?;
+        }
+
+        Ok(())
+    }
+
+    async fn schedule_relay_sync_retry(&self, relay_id: &str, source: &str) -> Result<()> {
+        let activities = self.query.list_activity_for_relay(relay_id).await?;
+        let consecutive_failures = consecutive_sync_failures(&activities);
+
+        let Some(delay) = relay_sync_retry_delay(consecutive_failures) else {
+            tracing::warn!(
+                relay = relay_id,
+                consecutive_failures,
+                max_attempts = RELAY_SYNC_RETRY_MAX_ATTEMPTS,
+                "relay sync retries exhausted; awaiting manual intervention"
+            );
+            return Ok(());
+        };
+
+        tracing::info!(
+            relay = relay_id,
+            source,
+            consecutive_failures,
+            delay_secs = delay.as_secs(),
+            "scheduled relay sync retry"
+        );
+
+        let relay_id = relay_id.to_string();
+        let infra = self.clone();
+
+        tokio::spawn(async move {
+            tokio::time::sleep(delay).await;
+
+            if let Err(e) = infra.retry_relay_sync(&relay_id).await {
+                tracing::error!(relay = %relay_id, error = %e, "relay sync retry task failed");
+            }
+        });
+
+        Ok(())
+    }
+
+    async fn retry_relay_sync(&self, relay_id: &str) -> Result<()> {
+        let Some(relay) = self.query.get_relay(relay_id).await? else {
+            return Ok(());
+        };
+
+        if relay.sync_error.trim().is_empty() {
+            tracing::debug!(relay = %relay.id, "skip relay sync retry; relay has no sync_error");
+            return Ok(());
+        }
+
+        let is_new = relay.synced == 0;
+        self.sync_and_report(&relay, is_new).await;
         Ok(())
     }
 
@@ -258,6 +336,28 @@ fn relay_sync_body(
 fn should_sync_relay_activity(activity_type: &str) -> bool {
     matches!(
         activity_type,
-        "create_relay" | "update_relay" | "activate_relay" | "deactivate_relay"
+        "create_relay" | "update_relay" | "activate_relay" | "deactivate_relay" | "fail_relay_sync"
     )
 }
+
+fn consecutive_sync_failures(activities: &[Activity]) -> usize {
+    activities
+        .iter()
+        .take_while(|activity| activity.activity_type == "fail_relay_sync")
+        .count()
+}
+
+fn relay_sync_retry_delay(consecutive_failures: usize) -> Option<Duration> {
+    let retry_attempt = consecutive_failures.max(1);
+    if retry_attempt > RELAY_SYNC_RETRY_MAX_ATTEMPTS {
+        return None;
+    }
+
+    let exponent = (retry_attempt - 1).min(31);
+    let multiplier = 1u64 << exponent;
+    let delay_secs = RELAY_SYNC_RETRY_BASE_DELAY_SECS
+        .saturating_mul(multiplier)
+        .min(RELAY_SYNC_RETRY_MAX_DELAY_SECS);
+
+    Some(Duration::from_secs(delay_secs))
+}

backend/src/query.rs

@@ -94,6 +94,23 @@ impl Query {
         Ok(rows)
     }
 
+    pub async fn list_relays_with_sync_error(&self) -> Result<Vec<Relay>> {
+        let rows = sqlx::query_as::<_, Relay>(
+            "SELECT id, tenant, schema, subdomain, plan, stripe_subscription_item_id,
+                     status, sync_error,
+                     info_name, info_icon, info_description,
+                     policy_public_join, policy_strip_signatures,
+                     groups_enabled, management_enabled, blossom_enabled,
+                     livekit_enabled, push_enabled, synced
+              FROM relay
+              WHERE TRIM(sync_error) != ''
+              ORDER BY id",
+        )
+        .fetch_all(&self.pool)
+        .await?;
+        Ok(rows)
+    }
+
     pub async fn list_relays_for_tenant(&self, tenant_id: &str) -> Result<Vec<Relay>> {
         let rows = sqlx::query_as::<_, Relay>(
             "SELECT id, tenant, schema, subdomain, plan, stripe_subscription_item_id,
@@ -144,6 +161,16 @@ impl Query {
         Ok(row)
     }
 
+    pub async fn get_invoice_nwc_payment_state(&self, invoice_id: &str) -> Result<Option<String>> {
+        let state = sqlx::query_scalar::<_, String>(
+            "SELECT state FROM invoice_nwc_payment WHERE invoice_id = ?",
+        )
+        .bind(invoice_id)
+        .fetch_optional(&self.pool)
+        .await?;
+        Ok(state)
+    }
+
     pub async fn has_active_paid_relays(&self, tenant_id: &str) -> Result<bool> {
         let plans = sqlx::query_scalar::<_, String>(
             "SELECT plan FROM relay WHERE tenant = ? AND status = 'active'",