I am trying to insert a compress string of size ~5Mb after compression as a byte array (Blob). when insert the actual size of the row is 23Mb which pass the database hard limit (16Mb).

I check the size of the array right before I send it to the driver. There could be an increase of size during the drive processing the data?

Thank you.

Consider the following code:

    let user = "cassandra";
    let password = "cassandra";
    let auth = StaticPasswordAuthenticatorProvider::new(&user, &password);
    let config = NodeTcpConfigBuilder::new()
        .with_contact_point("127.0.0.1:9042".into())
        .with_authenticator_provider(Arc::new(auth))
        .build()
        .await
        .unwrap();

    let session = TcpSessionBuilder::new(RoundRobinLoadBalancingStrategy::new(), config)
        .build()
        .unwrap();

    let mut event_recv = session.create_event_receiver();

    sleep(Duration::from_secs(3)).await; // let the driver finish connecting to the cluster and registering for the events

    let create_ks = "CREATE KEYSPACE IF NOT EXISTS test_events_ks WITH REPLICATION = { 'class' : 'SimpleStrategy', 'replication_factor' : 1 };";

    session.query(create_ks).await.unwrap();

    let event = timeout(Duration::from_secs(10), event_recv.recv())
        .await
        .unwrap()
        .unwrap();

Before https://github.com/krojew/cdrs-tokio/commit/bc021f4ee93c87bef922f5bf6cbac7b9780c10f6 event_recv.recv() would complete, with that commit it now times out

Hello there!

Prepared statement creating only in single node scope. With fixed nodes observablilty and enabled load balancing it become cause of query error:

ERROR cdrs_tokio::transport: Transport error! error=Server error: ErrorBody { error_code: 9472, message: "Prepared query with ID dde1cb1db179403026499e16ff7ef1f5 not found (either the query was not prepared on this host (maybe the host has been restarted?) or you have prepared too many queries and it has been evicted from the internal cache)", additional_info:

I recently came across https://github.com/scylladb/scylla-rust-driver And also saw https://www.reddit.com/r/rust/comments/pfuwhf/help_wanted_cdrstokio_road_to_performance_and/ So it seems you have seen scylla-rust-driver and that spurred a lot of the cleanup of cdrs's performance mistakes.

After evaluating the performance of scylla-rust-driver it still seems to considerably outperform master cdrs-tokio, in both full scale usage and protocol serializing/deserializing.

But you clearly seem dedicated to cdrs-tokio instead of jumping over to scylla-rust-driver, so what do you think cdrs-tokio does better than scylla-rust-driver in order for you to keep working on it?

I am getting error UnexpectedWriteType("CAS")) when doing batch inserts from multiple tasks in the same Session object.

It fails intermittently on a three node cluster with Cassandra version [cqlsh 6.0.0 | Cassandra 4.0.5 | CQL spec 3.4.5 | Native protocol v5]. It works consistently with a single task.

On a local Cassandra instance in docker, it work consistently with multiple tasks - version 4.0.6.

It seems this projects fork predates cassandra-proto

I am in need of a cassandra protocol crate that can support response encoding and query decoding for the purposes of implementing a proxy. Would that be out of scope of cdrs-tokio's protocol functionality? Would you be interested in pulling cdrs-tokio's protocol functionality into a cassandra-proto like crate? (I am happy to put in the work to make this happen)

I am currently looking at getting cassandra-proto cleaned up and improved at https://github.com/shotover/cassandra-proto But this crate is fairly active so it would be a shame to duplicate work. I would be happy to collaborate in any way that makes sense for your project: possibly a subcrate in this repo or possibly another separate repo.

for aws keyspaces, the authentication challenge includes a nonce key that is required for signing the auth token (when using aws credentials for cassandra/keyspaces access)

I think the most straightforward way to do this is to pass the auth challenge response body into Authenticator.auth_token() -- you can see from the aws sigv4 java plugin that this is how the datastax driver works for it's auth plugins

closes https://github.com/krojew/cdrs-tokio/issues/40

I tried to keep this PR as simple as possible, if there are other cleanups needed lets put them in a follow up PR. (Unless they make this PR simpler of course)

Hello. I am not sure if this is a bug or user error. When I am running my test suite against a database, I have noticed that at least a couple of the tests fairly consistently fail with an error: "Connection closed while waiting for response". I have tried to run the tests in single threaded mode, I have tried adding connections to the connection pool, but nothing seems to work. The number of failures vary, the tests which fail vary, etc. Since it also happens when running with "--jobs 1", I suspect it is not a threading issue.

At the same time, I see cassandra logs that say things like "Unknown exception in client networking" and "Connection reset by peer". I am running against a single node locally running cassandra test cluster.

I guess my question is whether this seems to be a bug or whether I am supposed to be handling intermittent connection failures like this in my client code.

Looking at the cassandra protocol spec and our current usage of frame.stream it makes a lot of sense for stream_id to be in Frame. What was the reasoning behind moving it out of Frame?

I have a static string reference in my struct, when I try to compile this piece of code the rust compiler errors with stackoverflow. If having a reference is not supported it should give warning or error instead of expanding the macro to use a code that makes rustc crash.

#[derive(IntoCdrsValue, TryFromRow, Clone, Debug, Serialize, Deserialize)]
pub struct EventData {
    pub task_id: String,
    pub agent: String,
    pub event_name: &'static str,
    pub event_time: u32,
    pub interaction_id: String,
}

When I convert the above event_name to a owned type such as String, it is able to complete the code.

Additional information

rustc 1.68.0-nightly (270c94e48 2022-12-28)

async-fn-in-trait is enabled. M1 Pro, ARM processor

When a SessionBuilder (e.g. TcpSessionBuilder) build() method is called, a Session is returned. However there are still background tasks that need to complete before the session can actually be used. Two examples I have hit are:

1. Delayed enabling of event processing - causing the broadcast::receiver returned by create_event_receiver to miss events
2. Delayed population of ClusterMetadataManager - causing failure to send prepared statement Cannot find node 127.0.0.1:9042 for statement re-preparation!

I think a reasonable solution to this would look like adding a tokio::sync::oneshot that gets sent when the session tasks have finished initializing and then the build() method can await on it.

This can be worked around locally by adding sleeps, but that is clearly not ideal.

Hello 👋 Is there any way to get tracing informations when making cassandra queries with this crate? I would have hope to get various spans showing the cassandra connections and query executions but I don't find any way to get it. Am I missing something or is this feature not available as of now ?

I was wondering if there is any way of using paging as described here: https://docs.datastax.com/en/developer/java-driver/4.14/manual/core/paging/ ? I am trying to have a select query, which returns many rows, combined to about 60Mb. the performance is slow (~0.8 sec). I think this is a paging issue.

let response = session
        .query_with_params(
            SELECT_CQL,
            StatementParamsBuilder::new()
                .with_values(query_values!(id, month))
                .build(),
        )
        .await?
        .response_body()?
        .into_rows();