Remove protocol versions >5 from MAX_SUPPORTED

[mykaul]

Without removing associated code yet, this patch just ensure we begin negotiating protocol version 5, then fallback to 4, reducing the amount of protocol errors that the Scylla server sends when it sees the unsupported versions.

Refs: #244

Perhaps we should add documentation to this?

Pre-review checklist

• I have split my patch into logically separate commits.
• All commit messages clearly explain what they change and why.
• I added relevant tests for new features and bug fixes.
• All commits compile, pass static checks and pass test.
• PR description sums up the changes and reasons why they should be introduced.
• I have provided docstrings for the public items that I want to introduce.
• I have adjusted the documentation in ./docs/source/.
• I added appropriate Fixes: annotations to PR description.

[mykaul]

Need to handle this failure:

=================================== FAILURES ===================================
  ________________ ConnectionTest.test_use_requested_compression _________________
  
  self = <tests.unit.test_connection.ConnectionTest testMethod=test_use_requested_compression>
  args = (), c = <Connection(139701745826928) 1.2.3.4:9042>
  options_buf = <_io.BytesIO object at 0x7f0ed8eb3c70>
  options = b'\x00\x02\x00\x0bCQL_VERSION\x00\x01\x00\x053.0.3\x00\x0bCOMPRESSION\x00\x02\x00\x06snappy\x00\x03lz4'
  
      def test_use_requested_compression(self, *args):
          c = self.make_connection()
          c._requests = {0: (c._handle_options_response, ProtocolHandler.decode_message, [])}
          c.defunct = Mock()
          # request snappy compression
          c.compression = "snappy"
      
          locally_supported_compressions.pop('lz4', None)
          locally_supported_compressions.pop('snappy', None)
          locally_supported_compressions['lz4'] = ('lz4compress', 'lz4decompress')
          locally_supported_compressions['snappy'] = ('snappycompress', 'snappydecompress')
      
          # the server only supports snappy
          options_buf = BytesIO()
          write_stringmultimap(options_buf, {
              'CQL_VERSION': ['3.0.3'],
              'COMPRESSION': ['snappy', 'lz4']
          })
          options = options_buf.getvalue()
      
          c.process_msg(_Frame(version=4, flags=0, stream=0, opcode=SupportedMessage.opcode, body_offset=9, end_pos=9 + len(options)), options)
      
  >       self.assertEqual(c.decompressor, locally_supported_compressions['snappy'][1])
  E       AssertionError: None != 'snappydecompress'
  
  /project/tests/unit/test_connection.py:216: AssertionError
  ------------------------------ Captured log call -------------------------------
  2025-06-19 14:32:29.361 DEBUG [protocol_features:62]: Parsing sharding info from message options {'COMPRESSION': ['snappy', 'lz4']}
  2025-06-19 14:32:29.361 DEBUG [connection:1352]: Received options response on new connection (139701745826928) from 1.2.3.4:9042
  2025-06-19 14:32:29.362 DEBUG [connection:1403]: Snappy compression is not supported with protocol version 5 and checksumming. Consider installing lz4. Disabling compression.
  2025-06-19 14:32:29.362 DEBUG [connection:1417]: Sending StartupMessage on <Connection(139701745826928) 1.2.3.4:9042>

[mykaul]

In #493 I've suggested (and implemented) removing protocols v1, v2. Here we need to decide if we wish to remove the higher versions.

[Lorak-mmk]

In #493 I've suggested (and implemented) removing protocols v1, v2. Here we need to decide if we wish to remove the higher versions.

I'm fine with removing DSE protocols, and all DSE functionality in general.

[Copilot]

Pull Request Overview

This PR restricts the driver to native protocol versions 3–5, makes V5 the default, and cleans up all DSE- and beta-protocol-specific paging logic and tests.

• Limit supported protocol versions to V3–V5 and set default to V5.
• Remove DSE/backpressured continuous-paging “next_pages” and queue-size behavior.
• Update and prune unit/integration tests to drop references to removed protocols and adapt compression negotiation.

Reviewed Changes

Copilot reviewed 20 out of 20 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
cassandra/init.py	Trim SUPPORTED_VERSIONS to (V5, V4, V3) and update docstring
cassandra/cluster.py	Change protocol_version default to V5; remove DSE‐paging support
cassandra/protocol.py	Drop writing of queue size and backpressure support
tests/unit/test_protocol.py	Simplify continuous-paging tests to only check unsupported cases
tests/unit/test_connection.py	Adapt compression negotiation to protocol V5 behavior
tests/integration/init.py	Update default/supported/unsupported protocol lists
tests/integration/*	Prune DSE‐ and protocol-6 tests and related decorators

Comments suppressed due to low confidence (2)

cassandra/protocol.py:617

• Since the code for writing max_queue_size has been removed, consider adding a unit test to verify that under protocol V5 only max_pages and max_pages_per_second are written and no queue-size field is emitted.

        write_int(f, paging_options.max_pages_per_second)

cassandra/init.py:175

• [nitpick] The docstring is now misleading; it should be updated to accurately describe the current supported versions (V3–V5) rather than referring to “future v5 support.”

    A tuple of all supported protocol versions for ScyllaDB, including future v5 support.

[mykaul]

Ugh. Wrong rebase.

[mykaul]

TODO: need to remove this continuous paging support.

[mykaul]

TODO: need to remove this continuous paging support.

Done, to some extent. I think it's reasonable.

[mykaul]

Flakiness?

 ================================== FAILURES ===================================
  ________________ TestTwistedTimer.test_multi_timer_validation _________________
  
  self = <tests.unit.io.test_twistedreactor.TestTwistedTimer testMethod=test_multi_timer_validation>
  
      def test_multi_timer_validation(self):
          """
          Verify that timer timeouts are honored appropriately
          """
          # Tests timers submitted in order at various timeouts
          submit_and_wait_for_completion(self, self.create_timer, 0, 100, 1, 100)
          # Tests timers submitted in reverse order at various timeouts
  >       submit_and_wait_for_completion(self, self.create_timer, 100, 0, -1, 100)
  
  D:\a\python-driver\python-driver\tests\unit\io\utils.py:166: 
  _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
  D:\a\python-driver\python-driver\tests\unit\io\utils.py:131: in submit_and_wait_for_completion
      unit_test.assertAlmostEqual(callback.expected_wait, callback.get_wait_time(), delta=.15)
  E   AssertionError: 0.09 != 0.2443854808807373 within 0.15 delta (0.1543854808807373 difference)

[mykaul]

Rebased, passed CI. I think it's OK now.

[mykaul]

If someone can review this, that'd be great. Not urgent.

[mykaul]

If someone can review this, that'd be great. Not urgent.

Same request, still not urgent.
However, I do see we haven't released a new Python driver for ages and we've accumulated tens of patches since last release (3.29.3-scylla...master ) - it would be good to get this in before the/a release.

[nyh]

I think there's some problem with how these two patches are split. The first patch claims it doesn't remove code, and yet it does remove a bunch of "continuous paging" and "graph paging" (whatever the heck that is) from the code - which later seems to be the focus of the second patch. Can you perhaps re-split the patches to better fit the descriptions?

Why did you want to remove the continuous paging stuff? What have we got to lose by having in the driver some DSE-only code? I know we don't recommend anyone using it, but if in the future we'll ever want to merge back with the upstream driver it will be easier if we don't change random pieces or the driver.
Or have we already given up the hope of merging from or to the upstream?

[nyh]

I'm ok with this PR, but it should be reviewed (and merged) by the drivers developers.

[mykaul]

@scylladb/python-driver-maint - please review (not urgent).

[mykaul]

I'm tempted to remove some more code, such as making monitor_reporting_enabled False, and some more... Please merge before I start pushing more code changes.

[dkropachev]

Looks good, one question though