Mini Shell

# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License"). You
# may not use this file except in compliance with the License. A copy of
# the License is located at
#
# http://aws.amazon.com/apache2.0/
#
# or in the "license" file accompanying this file. This file is
# distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF
# ANY KIND, either express or implied. See the License for the specific
# language governing permissions and limitations under the License.
import time
import logging
import weakref

from botocore import xform_name
from botocore.exceptions import BotoCoreError, HTTPClientError, ConnectionError
from botocore.model import OperationNotFoundError
from botocore.utils import CachedProperty

logger = logging.getLogger(__name__)


class EndpointDiscoveryException(BotoCoreError):
    pass


class EndpointDiscoveryRequired(EndpointDiscoveryException):
    """ Endpoint Discovery is disabled but is required for this operation. """
    fmt = 'Endpoint Discovery is not enabled but this operation requires it.'


class EndpointDiscoveryRefreshFailed(EndpointDiscoveryException):
    """ Endpoint Discovery failed to the refresh the known endpoints. """
    fmt = 'Endpoint Discovery failed to refresh the required endpoints.'


def block_endpoint_discovery_required_operations(model, **kwargs):
    endpoint_discovery = model.endpoint_discovery
    if endpoint_discovery and endpoint_discovery.get('required'):
        raise EndpointDiscoveryRequired()


class EndpointDiscoveryModel(object):
    def __init__(self, service_model):
        self._service_model = service_model

    @CachedProperty
    def discovery_operation_name(self):
        discovery_operation = self._service_model.endpoint_discovery_operation
        return xform_name(discovery_operation.name)

    @CachedProperty
    def discovery_operation_keys(self):
        discovery_operation = self._service_model.endpoint_discovery_operation
        keys = []
        if discovery_operation.input_shape:
            keys = list(discovery_operation.input_shape.members.keys())
        return keys

    def discovery_required_for(self, operation_name):
        try:
            operation_model = self._service_model.operation_model(operation_name)
            return operation_model.endpoint_discovery.get('required', False)
        except OperationNotFoundError:
            return False

    def discovery_operation_kwargs(self, **kwargs):
        input_keys = self.discovery_operation_keys
        # Operation and Identifiers are only sent if there are Identifiers
        if not kwargs.get('Identifiers'):
            kwargs.pop('Operation', None)
            kwargs.pop('Identifiers', None)
        return dict((k, v) for k, v in kwargs.items() if k in input_keys)

    def gather_identifiers(self, operation, params):
        return self._gather_ids(operation.input_shape, params)

    def _gather_ids(self, shape, params, ids=None):
        # Traverse the input shape and corresponding parameters, gathering
        # any input fields labeled as an endpoint discovery id
        if ids is None:
            ids = {}
        for member_name, member_shape in shape.members.items():
            if member_shape.metadata.get('endpointdiscoveryid'):
                ids[member_name] = params[member_name]
            elif member_shape.type_name == 'structure' and member_name in params:
                self._gather_ids(member_shape, params[member_name], ids)
        return ids


class EndpointDiscoveryManager(object):
    def __init__(self, client, cache=None, current_time=None, always_discover=True):
        if cache is None:
            cache = {}
        self._cache = cache
        self._failed_attempts = {}
        if current_time is None:
            current_time = time.time
        self._time = current_time
        self._always_discover = always_discover

        # This needs to be a weak ref in order to prevent memory leaks on
        # python 2.6
        self._client = weakref.proxy(client)
        self._model = EndpointDiscoveryModel(client.meta.service_model)

    def _parse_endpoints(self, response):
        endpoints = response['Endpoints']
        current_time = self._time()
        for endpoint in endpoints:
            cache_time = endpoint.get('CachePeriodInMinutes')
            endpoint['Expiration'] = current_time + cache_time * 60
        return endpoints

    def _cache_item(self, value):
        if isinstance(value, dict):
            return tuple(sorted(value.items()))
        else:
            return value

    def _create_cache_key(self, **kwargs):
        kwargs = self._model.discovery_operation_kwargs(**kwargs)
        return tuple(self._cache_item(v) for k, v in sorted(kwargs.items()))

    def gather_identifiers(self, operation, params):
        return self._model.gather_identifiers(operation, params)

    def delete_endpoints(self, **kwargs):
        cache_key = self._create_cache_key(**kwargs)
        if cache_key in self._cache:
            del self._cache[cache_key]

    def _describe_endpoints(self, **kwargs):
        # This is effectively a proxy to whatever name/kwargs the service
        # supports for endpoint discovery.
        kwargs = self._model.discovery_operation_kwargs(**kwargs)
        operation_name = self._model.discovery_operation_name
        discovery_operation = getattr(self._client, operation_name)
        logger.debug('Discovering endpoints with kwargs: %s', kwargs)
        return discovery_operation(**kwargs)

    def _get_current_endpoints(self, key):
        if key not in self._cache:
            return None
        now = self._time()
        return [e for e in self._cache[key] if now < e['Expiration']]

    def _refresh_current_endpoints(self, **kwargs):
        cache_key = self._create_cache_key(**kwargs)
        try:
            response = self._describe_endpoints(**kwargs)
            endpoints = self._parse_endpoints(response)
            self._cache[cache_key] = endpoints
            self._failed_attempts.pop(cache_key, None)
            return endpoints
        except (ConnectionError, HTTPClientError):
            self._failed_attempts[cache_key] = self._time() + 60
            return None

    def _recently_failed(self, cache_key):
        if cache_key in self._failed_attempts:
            now = self._time()
            if now < self._failed_attempts[cache_key]:
                return True
            del self._failed_attempts[cache_key]
        return False

    def _select_endpoint(self, endpoints):
        return endpoints[0]['Address']

    def describe_endpoint(self, **kwargs):
        operation = kwargs['Operation']
        discovery_required = self._model.discovery_required_for(operation)

        if not self._always_discover and not discovery_required:
            # Discovery set to only run on required operations
            logger.debug(
                'Optional discovery disabled. Skipping discovery for Operation: %s'
                % operation
            )
            return None

        # Get the endpoint for the provided operation and identifiers
        cache_key = self._create_cache_key(**kwargs)
        endpoints = self._get_current_endpoints(cache_key)
        if endpoints:
            return self._select_endpoint(endpoints)
        # All known endpoints are stale
        recently_failed = self._recently_failed(cache_key)
        if not recently_failed:
            # We haven't failed to discover recently, go ahead and refresh
            endpoints = self._refresh_current_endpoints(**kwargs)
            if endpoints:
                return self._select_endpoint(endpoints)
        # Discovery has failed recently, do our best to get an endpoint
        logger.debug('Endpoint Discovery has failed for: %s', kwargs)
        stale_entries = self._cache.get(cache_key, None)
        if stale_entries:
            # We have stale entries, use those while discovery is failing
            return self._select_endpoint(stale_entries)
        if discovery_required:
            # It looks strange to be checking recently_failed again but,
            # this informs us as to whether or not we tried to refresh earlier
            if recently_failed:
                # Discovery is required and we haven't already refreshed
                endpoints = self._refresh_current_endpoints(**kwargs)
                if endpoints:
                    return self._select_endpoint(endpoints)
            # No endpoints even refresh, raise hard error
            raise EndpointDiscoveryRefreshFailed()
        # Discovery is optional, just use the default endpoint for now
        return None


class EndpointDiscoveryHandler(object):
    def __init__(self, manager):
        self._manager = manager

    def register(self, events, service_id):
        events.register(
            'before-parameter-build.%s' % service_id, self.gather_identifiers
        )
        events.register_first(
            'request-created.%s' % service_id, self.discover_endpoint
        )
        events.register('needs-retry.%s' % service_id, self.handle_retries)

    def gather_identifiers(self, params, model, context, **kwargs):
        endpoint_discovery = model.endpoint_discovery
        # Only continue if the operation supports endpoint discovery
        if endpoint_discovery is None:
            return
        ids = self._manager.gather_identifiers(model, params)
        context['discovery'] = {'identifiers': ids}

    def discover_endpoint(self, request, operation_name, **kwargs):
        ids = request.context.get('discovery', {}).get('identifiers')
        if ids is None:
            return
        endpoint = self._manager.describe_endpoint(
            Operation=operation_name, Identifiers=ids
        )
        if endpoint is None:
            logger.debug('Failed to discover and inject endpoint')
            return
        if not endpoint.startswith('http'):
            endpoint = 'https://' + endpoint
        logger.debug('Injecting discovered endpoint: %s', endpoint)
        request.url = endpoint

    def handle_retries(self, request_dict, response, operation, **kwargs):
        if response is None:
            return None

        _, response = response
        status = response.get('ResponseMetadata', {}).get('HTTPStatusCode')
        error_code = response.get('Error', {}).get('Code')
        if status != 421 and error_code != 'InvalidEndpointException':
            return None

        context = request_dict.get('context', {})
        ids = context.get('discovery', {}).get('identifiers')
        if ids is None:
            return None

        # Delete the cached endpoints, forcing a refresh on retry
        # TODO: Improve eviction behavior to only evict the bad endpoint if
        # there are multiple. This will almost certainly require a lock.
        self._manager.delete_endpoints(
            Operation=operation.name, Identifiers=ids
        )
        return 0