transport::connection::tests::test_coalescing failed in CI

[Lorak-mmk]

https://github.com/scylladb/scylla-rust-driver/pull/863#issuecomment-1820496090

failures:

---- transport::connection::tests::test_coalescing stdout ----
Unique name: test_rust_1700493980_11
thread 'transport::connection::tests::test_coalescing' panicked at scylla/src/transport/connection.rs:2034:74:
called `Result::unwrap()` on an `Err` value: DbError(Unprepared { statement_id: b"i\xb60\x01/\xdc\x02\xb6\xec;\x18\x8e\xaa\xf38\xcf" }, "Prepared query with ID 69b630012fdc02b6ec3b188eaaf338cf 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)")

Strange; I know that we now sometimes prepare statements in case of query as well, but it falls back to execute which automatically reprepares the statement (but just once).

This test failed in CI, but the failure is unrelated to the PR it failed it. Needs to be investigated.

[piodul]

Fortunately, it's pretty easy to reproduce locally:

1. Modify the docker-compose command to use the test/cluster/cassandra/docker-compose.yml file,
2. make up to start the cluster
3. SCYLLA_URI=172.42.0.2:9042 SCYLLA_URI2=172.42.0.3:9042 SCYLLA_URI3=172.42.0.4:9042 cargo test -- test_coalescing

and it reproduces pretty much every time. What's interesting is that it's not necessary to run all the tests, running just this one test alone is sufficient to reproduce.

[roydahan]

Is this test still failing? flakey? or it was removed from CI?

[wprzytula]

It has recently failed: https://github.com/scylladb/scylla-rust-driver/actions/runs/9693468038/job/26748974433#step:6:1652

[muzarski]

Failed again: https://github.com/scylladb/scylla-rust-driver/actions/runs/10178379652/job/28151883467#step:6:1234