chore: increase retry time to dynamodb

rust/lance-table/src/io/commit.rs

@@ -678,7 +678,7 @@ async fn build_dynamodb_external_store(
 ) -> Result<Arc<dyn ExternalManifestStore>> {
     use super::commit::dynamodb::DynamoDBExternalManifestStore;
     use aws_sdk_dynamodb::{
-        config::{IdentityCache, Region},
+        config::{retry::RetryConfig, IdentityCache, Region},
         Client,
     };
 
@@ -687,7 +687,10 @@ async fn build_dynamodb_external_store(
         .region(Some(Region::new(region.to_string())))
         .credentials_provider(OSObjectStoreToAwsCredAdaptor(creds))
         // caching should be handled by passed AwsCredentialProvider
-        .identity_cache(IdentityCache::no_cache());
+        .identity_cache(IdentityCache::no_cache())
+        // Be more resilient to transient network issues.
+        // 5 attempts = 1 initial + 4 retries with exponential backoff.
+        .retry_config(RetryConfig::standard().with_max_attempts(5));
 
     if let Some(endpoint) = endpoint {
         dynamodb_config = dynamodb_config.endpoint_url(endpoint);

rust/lance-table/src/io/commit/dynamodb.rs

@@ -21,6 +21,7 @@ use object_store::path::Path;
 use snafu::location;
 use snafu::OptionExt;
 use tokio::sync::RwLock;
+use tracing::warn;
 
 use crate::io::commit::external_manifest::ExternalManifestStore;
 use lance_core::error::box_error;
@@ -82,7 +83,14 @@ where
     E: std::error::Error + Send + Sync + 'static,
 {
     fn wrap_err(self) -> Result<T> {
-        self.map_err(|err| Error::from(WrappedSdkError(err)))
+        self.map_err(|err| {
+            warn!(
+                target: "lance::dynamodb",
+                request_id = err.request_id().unwrap_or("unknown"),
+                "DynamoDB SDK error: {err:?}",
+            );
+            Error::from(WrappedSdkError(err))
+        })
     }
 }