Retry block_pointer_from_number on deploy (#4812)

Add retry to avoid a single failure while making the call prevent the subsgraph from being deployed
Add warning for failed attempts

core/src/subgraph/registrar.rs

@@ -13,6 +13,9 @@ use graph::prelude::{
     CreateSubgraphResult, SubgraphAssignmentProvider as SubgraphAssignmentProviderTrait,
     SubgraphRegistrar as SubgraphRegistrarTrait, *,
 };
+use graph::tokio_retry::Retry;
+use graph::util::futures::retry_strategy;
+use graph::util::futures::RETRY_DEFAULT_LIMIT;
 
 pub struct SubgraphRegistrar<P, S, SM> {
     logger: Logger,
@@ -518,15 +521,18 @@ async fn resolve_start_block(
         .expect("cannot identify minimum start block because there are no data sources")
     {
         0 => Ok(None),
-        min_start_block => chain
-            .block_pointer_from_number(logger, min_start_block - 1)
-            .await
-            .map(Some)
-            .map_err(move |_| {
-                SubgraphRegistrarError::ManifestValidationError(vec![
-                    SubgraphManifestValidationError::BlockNotFound(min_start_block.to_string()),
-                ])
-            }),
+        min_start_block => Retry::spawn(retry_strategy(Some(2), RETRY_DEFAULT_LIMIT), move || {
+            chain
+                .block_pointer_from_number(&logger, min_start_block - 1)
+                .inspect_err(move |e| warn!(&logger, "Failed to get block number: {}", e))
+        })
+        .await
+        .map(Some)
+        .map_err(move |_| {
+            SubgraphRegistrarError::ManifestValidationError(vec![
+                SubgraphManifestValidationError::BlockNotFound(min_start_block.to_string()),
+            ])
+        }),
     }
 }
 

graph/src/lib.rs

@@ -53,6 +53,7 @@ pub use semver;
 pub use slog;
 pub use stable_hash_legacy;
 pub use tokio;
+pub use tokio_retry;
 pub use tokio_stream;
 pub use url;
 

graph/src/util/futures.rs

@@ -9,6 +9,8 @@ use thiserror::Error;
 use tokio_retry::strategy::{jitter, ExponentialBackoff};
 use tokio_retry::Retry;
 
+pub const RETRY_DEFAULT_LIMIT: Duration = Duration::from_secs(30);
+
 /// Generic helper function for retrying async operations with built-in logging.
 ///
 /// To use this helper, do the following:
@@ -271,7 +273,7 @@ where
 
     let mut attempt_count = 0;
 
-    Retry::spawn(retry_strategy(limit_opt), move || {
+    Retry::spawn(retry_strategy(limit_opt, RETRY_DEFAULT_LIMIT), move || {
         let operation_name = operation_name.clone();
         let logger = logger.clone();
         let condition = condition.clone();
@@ -347,11 +349,17 @@ where
     })
 }
 
-fn retry_strategy(limit_opt: Option<usize>) -> Box<dyn Iterator<Item = Duration> + Send> {
+pub fn retry_strategy(
+    limit_opt: Option<usize>,
+    max_delay: Duration,
+) -> Box<dyn Iterator<Item = Duration> + Send> {
     // Exponential backoff, but with a maximum
-    let max_delay_ms = 30_000;
     let backoff = ExponentialBackoff::from_millis(2)
-        .max_delay(Duration::from_millis(max_delay_ms))
+        .max_delay(Duration::from_millis(
+            // This should be fine, if the value is too high it will crash during
+            // testing.
+            max_delay.as_millis().try_into().unwrap(),
+        ))
         .map(jitter);
 
     // Apply limit (maximum retry count)