Class RetryPolicy

cassandra/policies.py:714–875

A policy that describes whether to retry, rethrow, or ignore coordinator timeout and unavailable failures. These are failures reported from the server side. Timeouts are configured by `settings in cassandra.yaml <https://github.com/apache/cassandra/blob/cassandra-2.1.4/conf/cassandr

class RetryPolicy(object):
    """
    A policy that describes whether to retry, rethrow, or ignore coordinator
    timeout and unavailable failures. These are failures reported from the
    server side. Timeouts are configured by
    `settings in cassandra.yaml <https://github.com/apache/cassandra/blob/cassandra-2.1.4/conf/cassandra.yaml#L568-L584>`_.
    Unavailable failures occur when the coordinator cannot achieve the consistency
    level for a request. For further information see the method descriptions
    below.

    To specify a default retry policy, set the
    :attr:`.Cluster.default_retry_policy` attribute to an instance of this
    class or one of its subclasses.

    To specify a retry policy per query, set the :attr:`.Statement.retry_policy`
    attribute to an instance of this class or one of its subclasses.

    If custom behavior is needed for retrying certain operations,
    this class may be subclassed.
    """

    RETRY = 0
    """
    This should be returned from the below methods if the operation
    should be retried on the same connection.
    """

    RETHROW = 1
    """
    This should be returned from the below methods if the failure
    should be propagated and no more retries attempted.
    """

    IGNORE = 2
    """
    This should be returned from the below methods if the failure
    should be ignored but no more retries should be attempted.
    """

    RETRY_NEXT_HOST = 3
    """
    This should be returned from the below methods if the operation
    should be retried on another connection.
    """

    def on_read_timeout(self, query, consistency, required_responses,
                        received_responses, data_retrieved, retry_num):
        """
        This is called when a read operation times out from the coordinator&#x27;s
        perspective (i.e. a replica did not respond to the coordinator in time).
        It should return a tuple with two items: one of the class enums (such
        as :attr:`.RETRY`) and a :class:`.ConsistencyLevel` to retry the
        operation at or :const:`None` to keep the same consistency level.

        `query` is the :class:`.Statement` that timed out.

        `consistency` is the :class:`.ConsistencyLevel` that the operation was
        attempted at.