Unable to get transport in multi threaded application

[apezel]

Hi, thank you for your great work on CDRS, it's amazing.
I'm encountering some problem with the latest beta in a multi threaded application :

thread 'tokio-runtime-worker-1' panicked at 'called `Result::unwrap()` on an `Err` value: General("Unable to get transport")', libcore/result.rs:1009:5
note: Run with `RUST_BACKTRACE=1` for a backtrace.

I'm using lazy_static to share the connection between threads :

lazy_static! {
  pub static ref CASSANDRA: Arc<ClusterSession> = {
    new_connection()
  };
}

pub type ClusterSession = Session<RoundRobinSync<TcpConnectionPool<NoneAuthenticator>>>;

pub fn new_connection() -> Arc<ClusterSession> {
  let node = NodeTcpConfigBuilder::new(super::CONF.cassandra_host.as_str(), NoneAuthenticator {}).build();
  let cluster_config = ClusterTcpConfig(vec![node]);
  Arc::new(new_lz4(&cluster_config, RoundRobinSync::new()).expect("session should be created"))
}

inside the thread :

do_something_with_cassandra(&CASSANDRA.clone())

The problem occurs specially if i'm running large batch inserts in parallel.

Am i doing something wrong or is it a bug with CDRS ?

[AlexPikalov]

@apezel thank your creating the issue and the PR.

The way you use it seems to be okay for me. Probably existing solution with try_lock misses some retry logic. Your proposal #231 seems to be completely valid solution, but I just want to ensure it doesn't lead to unexpected panics.

[apezel]

Thanks !

[AlexPikalov]

Thank you for creating the issue and providing a solution