feat(policy): add shard reconnection policies

Add abstract classes: `ShardReconnectionPolicy` and `ShardReconnectionScheduler`
And implementations:
`NoDelayShardReconnectionPolicy` - policy that represents old behavior
of having no delay and no concurrency restriction.
`NoConcurrentShardReconnectionPolicy` - policy that limits concurrent
reconnections to 1 per scope and introduces delay between reconnections
within the scope.

Author: dkropachev

cassandra/policies.py

@@ -12,20 +12,28 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import random
+import threading
+import time
+import weakref
+from abc import ABC, abstractmethod
 
 from collections import namedtuple
+from enum import Enum
 from functools import lru_cache
 from itertools import islice, cycle, groupby, repeat
 import logging
 from random import randint, shuffle
 from threading import Lock
 import socket
 import warnings
+from typing import TYPE_CHECKING, Callable, Any, List
 
 log = logging.getLogger(__name__)
 
 from cassandra import WriteType as WT
 
+if TYPE_CHECKING:
+    from cluster import Session
 
 # This is done this way because WriteType was originally
 # defined here and in order not to break the API.
@@ -864,6 +872,210 @@ def _add_jitter(self, value):
         return min(max(self.base_delay, delay), self.max_delay)
 
 
+class ShardReconnectionScheduler(ABC):
+    @abstractmethod
+    def schedule(
+            self,
+            host_id: str,
+            shard_id: int,
+            method: Callable[..., Any],
+            *args: List[...],
+            **kwargs: dict[Any, Any]) -> None:
+        raise NotImplementedError()
+
+
+class ShardReconnectionPolicy(ABC):
+    """
+    Base class for shard reconnection policies.
+
+    On `new_scheduler` instantiate a scheduler that behaves according to the policy
+    """
+
+    @abstractmethod
+    def new_scheduler(self, session: Session) -> ShardReconnectionScheduler:
+        raise NotImplementedError()
+
+
+class NoDelayShardReconnectionPolicy(ShardReconnectionPolicy):
+    """
+    A shard reconnection policy with no delay between attempts and no concurrency restrictions.
+    Ensures at most one pending reconnection per (host, shard) pair — any additional
+     reconnection attempts for the same (host, shard) are silently ignored.
+
+    On `new_scheduler` instantiate a scheduler that behaves according to the policy
+    """
+
+    def new_scheduler(self, session: Session) -> ShardReconnectionScheduler:
+        return _NoDelayShardReconnectionScheduler(session)
+
+
+class _NoDelayShardReconnectionScheduler(ShardReconnectionScheduler):
+    def __init__(self, session: Session):
+        self.session: Session = weakref.proxy(session)
+        self.already_scheduled: dict[str, bool] = {}
+
+    def _execute(
+            self,
+            scheduled_key: str,
+            method: Callable[..., Any],
+            *args: List[...],
+            **kwargs: dict[Any, Any]) -> None:
+        try:
+            method(*args, **kwargs)
+        finally:
+            self.already_scheduled[scheduled_key] = False
+
+    def schedule(
+            self,
+            host_id: str,
+            shard_id: int,
+            method: Callable[..., Any],
+            *args: List[...],
+            **kwargs: dict[Any, Any]) -> None:
+        scheduled_key = f'{host_id}-{shard_id}'
+        if self.already_scheduled.get(scheduled_key):
+            return
+
+        self.already_scheduled[scheduled_key] = True
+        if not self.session.is_shutdown:
+            self.session.submit(self._execute, scheduled_key, method, *args, **kwargs)
+
+
+class ShardReconnectionPolicyScope(Enum):
+    """
+    A scope for `ShardReconnectionPolicy`, in particular `NoConcurrentShardReconnectionPolicy`
+    """
+    Cluster = 0
+    Host = 1
+
+
+class NoConcurrentShardReconnectionPolicy(ShardReconnectionPolicy):
+    """
+    A shard reconnection policy that allows only one pending connection per scope, where scope could be `Host`, `Cluster`
+    For backoff it uses `ReconnectionPolicy`, when there is no more reconnections to scheduled backoff policy is reminded
+    For all scopes does not allow schedule multiple reconnections for same host+shard, it silently ignores attempts to do that.
+
+    On `new_scheduler` instantiate a scheduler that behaves according to the policy
+    """
+
+    def __init__(
+            self,
+            shard_reconnection_scope: ShardReconnectionPolicyScope,
+            reconnection_policy: ReconnectionPolicy,
+    ):
+        if not isinstance(shard_reconnection_scope, ShardReconnectionPolicyScope):
+            raise ValueError("shard_reconnection_scope must be a ShardReconnectionPolicyScope")
+        if not isinstance(reconnection_policy, ReconnectionPolicy):
+            raise ValueError("reconnection_policy must be a ReconnectionPolicy")
+        self.shard_reconnection_scope = shard_reconnection_scope
+        self.reconnection_policy = reconnection_policy
+
+    def new_scheduler(self, session: Session) -> ShardReconnectionScheduler:
+        return _NoConcurrentShardReconnectionScheduler(session, self.shard_reconnection_scope, self.reconnection_policy)
+
+
+class _ScopeBucket:
+    """
+    Holds information for a shard reconnection scope, schedules and executes reconnections.
+    """
+
+    def __init__(
+            self,
+            session: Session,
+            reconnection_policy: ReconnectionPolicy,
+    ):
+        self._items = []
+        self.session = session
+        self.reconnection_policy = reconnection_policy
+        self.lock = threading.Lock()
+        self.running = False
+        self.schedule = self.reconnection_policy.new_schedule()
+
+    def _get_delay(self):
+        if self.schedule is None:
+            self.schedule = self.reconnection_policy.new_schedule()
+        try:
+            return next(self.schedule)
+        except StopIteration:
+            self.schedule = self.reconnection_policy.new_schedule()
+            return next(self.schedule)
+
+    def _schedule(self):
+        if self.session.is_shutdown:
+            return
+        delay = self._get_delay()
+        if delay:
+            self.session.cluster.scheduler.schedule(delay, self._run)
+        else:
+            self.session.submit(self._run)
+
+    def _run(self):
+        if self.session.is_shutdown:
+            return
+
+        with self.lock:
+            try:
+                item = self._items.pop()
+            except IndexError:
+                self.running = False
+                self.schedule = None
+                return
+
+        method, args, kwargs = item
+        try:
+            method(*args, **kwargs)
+        finally:
+            self._schedule()
+
+    def add(self, method, *args, **kwargs):
+        with self.lock:
+            self._items.append([method, args, kwargs])
+            if not self.running:
+                self.running = True
+                self._schedule()
+
+
+class _NoConcurrentShardReconnectionScheduler(ShardReconnectionScheduler):
+    def __init__(
+            self,
+            session: Session,
+            shard_reconnection_scope: ShardReconnectionPolicyScope,
+            reconnection_policy: ReconnectionPolicy,
+    ):
+        self.already_scheduled: dict[str, bool] = {}
+        self.scopes: dict[str, _ScopeBucket] = {}
+        self.shard_reconnection_scope = shard_reconnection_scope
+        self.reconnection_policy = reconnection_policy
+        self.session = session
+        self.lock = threading.Lock()
+
+    def _execute(self, scheduled_key, method, *args, **kwargs):
+        try:
+            method(*args, **kwargs)
+        finally:
+            with self.lock:
+                self.already_scheduled[scheduled_key] = False
+
+    def schedule(self, host_id, shard_id, method, *args, **kwargs):
+        if self.shard_reconnection_scope == ShardReconnectionPolicyScope.Cluster:
+            scope_hash = "global-cluster-scope"
+        else:
+            scope_hash = host_id
+        scheduled_key = f'{host_id}-{shard_id}'
+
+        with self.lock:
+            if self.already_scheduled.get(scheduled_key):
+                return False
+            self.already_scheduled[scheduled_key] = True
+
+            scope_info = self.scopes.get(scope_hash, 0)
+            if not scope_info:
+                scope_info = _ScopeBucket(self.session, self.reconnection_policy)
+                self.scopes[scope_hash] = scope_info
+            scope_info.add(self._execute, scheduled_key, method, *args, **kwargs)
+            return True
+
+
 class RetryPolicy(object):
     """
     A policy that describes whether to retry, rethrow, or ignore coordinator