fix: Default kafka timeout to be shorter than gRPC timeout (60 sec --> 10 sec) (#3131)

* fix: Default kafka timeout to be shorter than gRPC timeout

* docs: fix link style

influxdb_iox/tests/end_to_end_cases/management_api.rs

@@ -8,6 +8,7 @@ use generated_types::{
 };
 use influxdb_iox_client::{
     management::{Client, CreateDatabaseError},
+    router::generated_types::{write_buffer_connection, WriteBufferConnection},
     write::WriteError,
 };
 use std::{fs::set_permissions, os::unix::fs::PermissionsExt};
@@ -184,6 +185,37 @@ async fn test_create_database_invalid_name() {
     assert!(matches!(dbg!(err), CreateDatabaseError::InvalidArgument(_)));
 }
 
+#[tokio::test]
+async fn test_create_database_invalid_kafka() {
+    let server_fixture = ServerFixture::create_shared(ServerType::Database).await;
+    let mut client = server_fixture.management_client();
+
+    let rules = DatabaseRules {
+        name: "db_with_bad_kafka_address".into(),
+        write_buffer_connection: Some(WriteBufferConnection {
+            direction: write_buffer_connection::Direction::Read.into(),
+            r#type: "kafka".into(),
+            connection: "i_am_not_a_kafka_server:1234".into(),
+            ..Default::default()
+        }),
+        ..Default::default()
+    };
+
+    let start = Instant::now();
+    let err = client
+        .create_database(rules)
+        .await
+        .expect_err("expected request to fail");
+
+    println!("Failed after {:?}", Instant::now() - start);
+
+    // expect that this error has a useful error related to kafka (not "timeout")
+    assert_contains!(
+        err.to_string(),
+        "error creating write buffer: Meta data fetch error: BrokerTransportFailure"
+    );
+}
+
 #[tokio::test]
 async fn test_list_databases() {
     let server_fixture = ServerFixture::create_shared(ServerType::Database).await;

write_buffer/src/kafka.rs

@@ -37,6 +37,16 @@ use crate::{
     },
 };
 
+/// Default timeout supplied to rdkafka client for kafka operations.
+///
+/// Chosen to be a value less than the default gRPC timeout (30
+/// seconds) so we can detect kafka errors and return them prior to
+/// the gRPC requests to IOx timing out.
+///
+/// More context in
+/// <https://github.com/influxdata/influxdb_iox/issues/3029>
+const KAFKA_OPERATION_TIMEOUT_MS: u64 = 10000;
+
 impl From<&IoxHeaders> for OwnedHeaders {
     fn from(iox_headers: &IoxHeaders) -> Self {
         let mut res = Self::new();
@@ -252,7 +262,7 @@ impl WriteBufferReading for KafkaBufferConsumer {
                         consumer_cloned.fetch_watermarks(
                             &database_name,
                             sequencer_id as i32,
-                            Duration::from_secs(60),
+                            Duration::from_millis(KAFKA_OPERATION_TIMEOUT_MS),
                         )
                     })
                     .await
@@ -299,7 +309,7 @@ impl WriteBufferReading for KafkaBufferConsumer {
                     &database_name,
                     sequencer_id as i32,
                     offset,
-                    Duration::from_secs(60),
+                    Duration::from_millis(KAFKA_OPERATION_TIMEOUT_MS),
                 )
             })
             .await
@@ -404,7 +414,10 @@ async fn get_partitions(
     let metadata = tokio::task::spawn_blocking(move || {
         let probe_consumer: BaseConsumer = cfg.create()?;
 
-        probe_consumer.fetch_metadata(Some(&database_name), Duration::from_secs(60))
+        probe_consumer.fetch_metadata(
+            Some(&database_name),
+            Duration::from_millis(KAFKA_OPERATION_TIMEOUT_MS),
+        )
     })
     .await
     .expect("subtask failed")?;