Add Session.wait_for_schema_agreement() as a session-scoped schema agreement check. The new API queries schema_version from system.local on the connected hosts selected by the requested rack, dc, or cluster scope, respects Cluster.max_schema_agreement_wait and the control-connection metadata timeouts, and bounds the fan-out with configurable parallelism.

Update the public Session docs and switch the integration callers that were explicitly waiting on schema agreement to use the session API. Add unit coverage for agreement, retries, busy connections, missing pools, batching, scope filtering, and invalid scope handling.

…ment

Keep ControlConnection.wait_for_schema_agreement() as a compatibility wrapper, but move the existing implementation to _wait_for_schema_agreement() and deprecate the public method in favor of Session.wait_for_schema_agreement(). This lets the control-connection refresh path continue using the old logic internally without emitting warnings.

The control-connection wait path was designed for internal metadata refresh use, not as a user-facing schema agreement API.
It observes schema agreement from one single node, assuming that schema change statement have been ran on that host.
Using it by users will lead to false positives, if user ran statement on
a host different from host of control connection.

Update the unit tests to call the internal helper everywhere a warning is not expected, add explicit deprecation coverage for the public wrapper, and set stacklevel=2 so the warning points at the caller instead of inside the driver.

Keep heartbeat request-id and in-flight bookkeeping consistent when send_msg() fails.\n\nHandle the control-connection in_flight release separately from HostConnection cleanup.

Add an opt-in control-connection fallback for application queries when the driver cannot populate normal node pools, which happens in deployments that expose the cluster through a non-broadcast IP address such as a TCP proxy or a node public IP. In that mode the driver can still execute queries over the single control connection, but throughput is poor and connection churn increases the chance of request errors. This option is intentionally disabled by default and should not be used in production.

Also propagate keyspace updates on the fallback path so USE keeps the control connection in sync.

Tests:
- tests/unit/test_cluster.py::ClusterTest::test_set_keyspace_for_all_pools_reports_all_errors
- tests/unit/test_response_future.py::ResponseFutureTests::test_control_connection_fallback_updates_connection_keyspace

Replace SimpleStrategy with NetworkTopologyStrategy across integration
tests to align with ScyllaDB's tablet-based replication defaults.

In the tablets test module, skip default keyspace creation
(set_keyspace=False) to avoid RF=3 keyspaces that block node
decommission when all nodes already hold replicas.

With tablets enabled, decommissioning a node from a 3-node cluster with
RF=3 fails because there is no available node to receive tablet replicas.
Bootstrap 3 replacement nodes instead of 2 so that each original node
can be decommissioned while sufficient replicas remain.

LWT is not supported with tablets on ScyllaDB < 2025.4. Mark the affected SerialConsistencyTests and
LightweightTransactionTests as xfail for those versions.

Secondary indexes are not supported on base tables with tablets for Scylla
versions < 2026.1.

…ts disabled

The integration test suite was pinned to release:2025.2 which is no
longer the latest LTS branch.  Update to release:2026.1 so CI covers
the newest ScyllaDB features and catches regressions earlier.

Tests gated by @skip_scylla_version_lt(2026.1.0) — such as the
client_routes tests — will now actually execute in CI.

ScyllaDB has dropped support for SimpleStrategy. Update all CQL
statements, test fixtures, examples, benchmarks, and management
utilities to use NetworkTopologyStrategy instead.

The SimpleStrategy class definition in cassandra/metadata.py is
preserved for backward compatibility with Cassandra clusters.

Bumps [urllib3](https://github.com/urllib3/urllib3) from 2.6.3 to 2.7.0.
- [Release notes](https://github.com/urllib3/urllib3/releases)
- [Changelog](https://github.com/urllib3/urllib3/blob/main/CHANGES.rst)
- [Commits](urllib3/urllib3@2.6.3...2.7.0)

---
updated-dependencies:
- dependency-name: urllib3
  dependency-version: 2.7.0
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>

…ylladb#22677

Tests that previously xfailed on ScyllaDB < 2026.1 due to MVs, secondary
indexes, and counters not being supported on tables with tablets now create
their keyspace with 'AND tablets = {"enabled": false}' for those older
versions, so the tests run and pass rather than being expected to fail.

A new helper get_tablets_disabled_ddl_suffix() is added to
tests/integration/__init__.py to return the appropriate DDL suffix.

Python 3.8+ rejects ssl.SSLSocket in asyncio's sock_sendall/sock_recv
with TypeError. This caused the driver to fail connecting to ScyllaDB
clusters requiring TLS, manifesting as 'protocol version 21 not
supported' errors (0x15 = TLS Alert byte misread as protocol version).

Fix by using asyncio's native TLS transport (loop.create_connection
with ssl= parameter) instead of wrapping sockets with
ssl.SSLContext.wrap_socket(). This preserves shard-aware port binding
done during _initiate_connection().

Add _AsyncioProtocol to bridge asyncio's transport/protocol API back
to Connection.process_io_buffer() for SSL data reads. Non-SSL
connections continue using the existing sock_recv path.

Fixes scylladb#330

- Update all action references to use full SHA commit hashes
- Configure Renovate to pin digests and require 90-day minimum age
- Add github-actions ecosystem to Dependabot