ryerraguntla commented on code in PR #4043: URL: https://github.com/apache/iggy/pull/4043#discussion_r3965851129
########## gateways/kafka/src/bridge/iggy_bridge.rs: ########## @@ -0,0 +1,310 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, +// software distributed under the License is distributed on an +// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +// KIND, either express or implied. See the License for the +// specific language governing permissions and limitations +// under the License. + +use iggy::prelude::{ + AutoLogin, Client, Credentials, Identifier, IggyClient, IggyClientBuilder, IggyError, + StreamClient, TopicClient, TopicCreateOptions, +}; +use tracing::{debug, info}; + +use crate::bridge::config::IggyBridgeConfig; +use crate::bridge::error::BridgeError; + +/// Passes attempted, after the first, before [`IggyBridge::connect`] gives up and returns `Err`. +/// +/// Not the SDK's own default (`TcpClientReconnectionConfig::default()` is `max_retries: None` - +/// unlimited, one dial per second, forever). A Kafka client already retries at the wire-protocol +/// level once a handler maps a bridge failure to a retriable error code; the bridge blocking a +/// request task inside an unbounded internal reconnect loop would just add a second, invisible +/// retry layer underneath that one instead of surfacing the failure so the mapped code can be +/// sent. At the default `reconnection_interval` (1s), a fully unreachable address fails in a few +/// seconds rather than hanging. +const RECONNECTION_RETRIES: u32 = 3; + +/// Owns one connected `IggyClient` and resolves Kafka topics against it. +/// +/// Produce/Fetch handler wiring is a separate, later change (`#3535`/`#3536`) - this type is the +/// shared plumbing those handlers will call into, exercised standalone here via its own tests and +/// an integration test against a real `iggy-server`. +pub struct IggyBridge { + client: IggyClient, + config: IggyBridgeConfig, +} + +impl IggyBridge { + /// Connects to Iggy using `config` and authenticates. + /// + /// Builds the client through the SDK's fluent TCP builder rather than hand-assembling an + /// `iggy://user:pass@host` connection string: that string format splits on `@` then `:`, so a + /// password containing either character (`p@ss:word`) would be misparsed into a garbled + /// address instead of failing with a diagnosable config error. The fluent builder passes + /// `username`/`password` as already-separated fields, sidestepping the ambiguity entirely. + /// + /// # Errors + /// + /// Returns [`BridgeError::InvalidConfig`] if `config.address` is empty. Returns + /// [`BridgeError::Iggy`] if the address is malformed, the TCP connection fails, or + /// authentication is rejected - this is the boundary [`BridgeError::to_kafka_error_code`] + /// exists for: a handler calling this must map the error to a wire response, never panic or + /// unwrap, since an unreachable Iggy backend is an expected runtime condition, not a bug. + pub async fn connect(config: IggyBridgeConfig) -> Result<Self, BridgeError> { + if config.address.trim().is_empty() { + return Err(BridgeError::InvalidConfig( + "Iggy address must not be empty".to_string(), + )); + } + + let credentials = + Credentials::UsernamePassword(config.username.clone(), config.password.clone()); + let client = IggyClientBuilder::new() + .with_tcp() + .with_server_address(config.address.clone()) + .with_auto_sign_in(AutoLogin::Enabled(credentials)) + .with_reconnection_max_retries(Some(RECONNECTION_RETRIES)) + .build() + .map_err(BridgeError::Iggy)?; + client.connect().await.map_err(BridgeError::Iggy)?; + info!("Iggy bridge connected to {}", config.address); + + Ok(Self { client, config }) + } + + /// Disconnects the underlying Iggy client. + /// + /// # Errors + /// + /// Returns [`BridgeError::Iggy`] if the client reports a disconnect failure. + pub async fn close(&self) -> Result<(), BridgeError> { + self.client.disconnect().await.map_err(BridgeError::Iggy) + } + + /// Ensures the Iggy stream and topic backing `kafka_topic` exist, creating either or both if + /// missing. Resolves `kafka_topic` through the configured [`TopicMapping`](crate::bridge::topic_map::TopicMapping). + /// + /// Idempotent when repeated with the *same* `partition_count`: a `get` before each `create` + /// means calling this twice for the same topic is a no-op the second time, and a + /// `NameAlreadyExists` race from a concurrent caller creating the same stream/topic between + /// this call's `get` and `create` is treated as success, not an error - the desired end state + /// (it exists) is what idempotency actually promises, not that this call was the one that + /// created it. A *different* `partition_count` against an already-existing topic is not + /// idempotent - see [`BridgeError::PartitionCountMismatch`]. + /// + /// # Errors + /// + /// Returns [`BridgeError::Iggy`] for connectivity/auth/invalid-name failures. Returns + /// [`BridgeError::PartitionCountMismatch`] if the topic already exists with a different + /// partition count than `partition_count`. + pub async fn ensure_stream_and_topic( + &self, + kafka_topic: &str, + partition_count: u32, + ) -> Result<(), BridgeError> { + let (stream_name, topic_name) = self.config.topic_mapping.resolve(kafka_topic); + let stream_id = self.ensure_stream(&stream_name).await?; + self.ensure_topic(&stream_id, &topic_name, kafka_topic, partition_count) + .await?; + Ok(()) + } + + /// Looks up (or creates) the stream named `stream_name`. + /// + /// `Identifier::named` - never `Identifier::try_from`/`FromStr` - because the latter parses + /// an all-digit string as a numeric Iggy ID rather than a name. A stream or topic named e.g. + /// `"42"` would otherwise resolve against the wrong resource on every call after the first: + /// the first `ensure_stream_and_topic("42", ...)` creates a stream *named* `"42"`, but a + /// second call would look it up *by ID* `42` instead, almost certainly finding nothing and + /// breaking the "idempotent on repeated calls" guarantee. + async fn ensure_stream(&self, stream_name: &str) -> Result<Identifier, BridgeError> { + let identifier = Identifier::named(stream_name).map_err(BridgeError::Iggy)?; + if let Some(existing) = self + .client + .get_stream(&identifier) + .await + .map_err(BridgeError::Iggy)? + { + debug!("Iggy stream '{stream_name}' already exists"); + return Identifier::numeric(existing.id).map_err(BridgeError::Iggy); Review Comment: The named-vs-numeric doc is doing the right thing on lookup, then throwing it away. Recycled stream ids after delete+recreate would send `create_topic` at the wrong stream. Will pass the `Identifier::named(stream_name)` already built at the start of `ensure_stream` into `ensure_topic`. That also drops the `Identifier::numeric` `map_err` calls (infallible for any `u32`) and stops printing a slab key in the `:226` not-found path. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
