Use Datastax Cassandra java driver v4

[findinpath]

Description

This PR addresses #7729

Upgrade to latest java-driver-core version 4.14.0 in order to make use
of the latest improvements in the Cassandra java driver.

The version 4 of the Cassandra java driver is not backwards compatible
with the version 3 although most of the concepts still overlap.

Most of the changes revolve to a very customizable way of building the CqlSession
(which is the replacement for Cluster & Session from the version 3 of the driver).
This is why there are a great deal of changes in the CassandraClientModule class.
Drop the support for WhiteListPolicy because this load balancing policy
is not anymore present in v4 of the Cassandra driver.

HostAddressFactory deals currently only with nodes with endpoints which
resolve to InetSocketAddress.

Keep working with internal String names for tables and schema instead of
the newly introduced CqlIdentifier in order to keep the compatibility
to trino-spi layer.

NOTE that these changes introduce breaking changes in the configuration of the
trino-cassandra connector for the following sections:

• load balancing policy
• retry policy type

Is this change a fix, improvement, new feature, refactoring, or other?

Refactoring

Is this a change to the core query engine, a connector, client library, or the SPI interfaces? (be specific)

trino-cassandra connector

How would you describe this change to a non-technical end user or system administrator?

The Datastax Cassandra Java driver is described like:

A modern, feature-rich and highly tunable Java client library for Apache Cassandra® (2.1+) and DataStax Enterprise (4.7+), and DataStax Astra, using exclusively Cassandra’s binary protocol and Cassandra Query Language (CQL) v3.

Pragmatically speaking, the driver for Cassandra 3.x won't get new functionality, only critical bug fixes. So it would be wise to switch to the actively maintained library.

Related issues, pull requests, and links

Fixes #7729

Once this PR is being merged and actively used in trino, the repository https://github.com/trinodb/trino-cassandra-driver can be archived because it will no longer be in use.

Documentation

(x) No documentation is needed.
( ) Sufficient documentation is included in this PR.
( ) Documentation PR is available with #prnumber.
( ) Documentation issue #issuenumber is filed, and can be handled later.

Release notes

( ) No release notes entries required.
(x) Release notes entries required with the following suggested text:

# Cassandra
* Use Datastax java driver v4 to communicate with Cassandra. 
  This change adds supports for v5 and v6 protocol and removes support for v2 protocol. Also, it makes 
  `cassandra.load-policy.use-dc-aware` and `cassandra.load-policy.dc-aware.local-dc` 
  config properties required.  ({issue}`7729`)

[findinpath]

This PR is an initial draft (although the trino-cassandra project builds and should have green tests) and has the purpose of getting feedback on how to shape the connector with Cassandra java driver v4 in order to be production ready.

[findinpath]

Useful documentation resources:

• https://docs.datastax.com/en/developer/java-driver/4.11/
• https://docs.datastax.com/en/developer/java-driver/3.11/
• https://github.com/datastax/java-driver
• https://github.com/DataStax-Examples/java-cassandra-driver-from3x-to4x
• https://github.com/datastax/java-driver/blob/4.x/core/src/main/resources/reference.conf -> programmatic configurations available for creating a CqlSession instance.

[kokosing]

Should we have split cassandra plugin into two like we did with phoenix, so we could have two plugins cassandra3 and cassadra5. Higher maintenance cost but no backward compatible changes. WDYT?

[hashhar]

The Cassandra version being dropped (2.0.x) has had no new releases since 2015.

The version 4 of the Cassandra java driver is not backwards compatible
with the version 3 although most of the concepts still overlap.

This refers to the driver version. The 4.x driver is compatible with Cassandra 2.1.x which is the oldest active release (released in 2014). (@findinpath It would help to clarify this in the commit message and PR description. Instead of talking about driver versions we can talk about Cassandra versions which is more clear.)

[findinpath]

@findinpath It would help to clarify this in the commit message and PR description. Instead of talking about driver versions we can talk about Cassandra versions which is more clear.

@hashhar I am pretty sure that this commit should probably be split in several other commits to provide more insights for the maintainer in why certain changes have been made. My initial intention was to have a decent draft on which we can start the discussion.

[findinpath]

Should we have split cassandra plugin into two like we did with phoenix, so we could have two plugins cassandra3 and cassadra5.

@kokosing
https://docs.datastax.com/en/developer/java-driver/4.11/

A modern, feature-rich and highly tunable Java client library for Apache Cassandra® (2.1+) and DataStax Enterprise (4.7+), using exclusively Cassandra’s binary protocol and Cassandra Query Language v3.

I am actually more in favor of doing the update of the cassandra client library and take advantage of the richness of configuration options for CqlSession builder.

[kokosing]

I understand that V2 no longer works, am I right? If so we should not allow users to set this.

[findinpath]

Indeed @kokosing . Dropping the support for V2 of the protocol was a design decision from Datastax

Please see also: #7729 (comment)

[findinpath]

The approach for setting the load balancing policy in the v4 version of the driver implies to rethink the appropriate settings for the cassandra trino plugin.

Have a look over https://github.com/datastax/java-driver/blob/4.x/core/src/main/resources/reference.conf for the available options for load balancing policy:

• DefaultLoadBalancingPolicy
• BasicLoadBalancingPolicy
• DcInferringLoadBalancingPolicy
• ...

NOTE that the current implementation hardcodes the usage of DefaultLoadBalancingPolicy load balancing policy.

[ebyhr]

@findinpath Sorry for my late review. Could you rebase on upstream to resolve conflicts?

[robd003]

@ebyhr Just did a rebase here: https://github.com/robd003/trino

[findinpath]

@robd003 thanks for the help.

@ebyhr concerning this PR, the tests were passing at the time when I wrote the PR, but I have identified at the time the fact that cassandra v4 driver offers much much more configuration capabilities compared to cassandra v3 driver.

See #7828 (comment)

The branch has now been rebased on top of master

[findinpath]

@ebyhr regarding the comments related to having more commits, I definitely agree with you.

Initially I considered this PR just a draft and wanted to first gather feedback regarding the implementation. Moreover, I wasn't sure whether this PR has any chance on landing on master.

I would split the commit in more commits as soon as we have a basic agreement on whether this PR makes sense to land on master.

After analysing the content of the commit, it seems that all the files changed are linked to upgrading the cassandra driver version and for this reason I think they should share the same commit. Please advice whether you see a better approach on how to split the commit in more readable / independent commits (which still compile and have the tests running successfully).

[findinpath]

Is here IAE appropriate or rather TrinoException ?

[ebyhr]

Could you update cassandra.rst where mentions protocol versions?

[findinpath]

Rebasing on master to solve compilation issue in TestCassandraTypeMapping.java

[findinpath]

@mosabua FYI