Merge pull request #16 from edx/clintonb/es-cmd-update

Updated install_es_indexes management command

course_discovery/apps/core/tests/mixins.py

@@ -3,7 +3,7 @@ import logging
 from django.conf import settings
 from elasticsearch import Elasticsearch
 
-from course_discovery.apps.courses.config import COURSES_INDEX_CONFIG
+from course_discovery.apps.courses.utils import ElasticsearchUtils
 
 logger = logging.getLogger(__name__)
 
@@ -23,17 +23,23 @@ class ElasticsearchTestMixin(object):
 
     def reset_index(self):
         """ Deletes and re-creates the Elasticsearch index. """
+        self.delete_index(self.index)
+        ElasticsearchUtils.create_alias_and_index(self.es, self.index)
 
-        index = self.index
+    def delete_index(self, index):
+        """
+        Deletes an index.
+
+        Args:
+            index (str): Name of index to delete
 
+        Returns:
+            None
+        """
         logger.info('Deleting index [%s]...', index)
         self.es.indices.delete(index=index, ignore=404)  # pylint: disable=unexpected-keyword-arg
         logger.info('...index deleted.')
 
-        logger.info('Recreating index [%s]...', index)
-        self.es.indices.create(index=index, body=COURSES_INDEX_CONFIG)
-        logger.info('...done!')
-
     def refresh_index(self):
         """
         Refreshes an index.

course_discovery/apps/courses/management/commands/install_es_indexes.py

@@ -2,31 +2,22 @@ import logging
 
 from django.conf import settings
 from django.core.management import BaseCommand
+from elasticsearch import Elasticsearch
 
-from elasticsearch import Elasticsearch, TransportError
-
-from course_discovery.apps.courses.config import COURSES_INDEX_CONFIG
+from course_discovery.apps.courses.utils import ElasticsearchUtils
 
 logger = logging.getLogger(__name__)
 
 
 class Command(BaseCommand):
-    help = 'Install any required elasticsearch indexes'
+    help = 'Install any required Elasticsearch indexes'
 
     def handle(self, *args, **options):
         host = settings.ELASTICSEARCH['host']
-        index = settings.ELASTICSEARCH['index']
+        alias = settings.ELASTICSEARCH['index']
 
         logger.info('Attempting to establish initial connection to Elasticsearch host [%s]...', host)
         es = Elasticsearch(host, sniff_on_start=True)
         logger.info('...success!')
 
-        logger.info('Making sure index [%s] exists...', index)
-        try:
-            es.indices.create(index=index, body=COURSES_INDEX_CONFIG)
-            logger.info('...index created.')
-        except TransportError as e:
-            if e.status_code == 400:
-                logger.info('...index already exists.')
-            else:
-                raise
+        ElasticsearchUtils.create_alias_and_index(es, alias)

course_discovery/apps/courses/tests/test_install_es_indexes.py

-import mock
 from django.conf import settings
-from django.test import TestCase
 from django.core.management import call_command
-from elasticsearch import TransportError
-from elasticsearch.client import IndicesClient
-from testfixtures import LogCapture
+from django.test import TestCase
 
 from course_discovery.apps.core.tests.mixins import ElasticsearchTestMixin
 
 LOGGER_NAME = 'courses.management.commands.install_es_indexes'
 
 
-class CourseInstallEsIndexes(ElasticsearchTestMixin, TestCase):
-    def test_ready_create_index(self):
-        """ Verify the app does not setup a new Elasticsearch index if one exists already. """
-        host = settings.ELASTICSEARCH['host']
+class InstallEsIndexesCommandTests(ElasticsearchTestMixin, TestCase):
+    def test_create_index(self):
+        """ Verify the app sets the alias and creates a new index. """
         index = settings.ELASTICSEARCH['index']
 
         # Delete the index
         self.es.indices.delete(index=index, ignore=404)  # pylint: disable=unexpected-keyword-arg
         self.assertFalse(self.es.indices.exists(index=index))
 
-        with LogCapture(LOGGER_NAME) as l:
-            call_command('install_es_indexes')
+        call_command('install_es_indexes')
 
-            # Verify the index was created
-            self.assertTrue(self.es.indices.exists(index=index))
-
-        l.check(
-            (LOGGER_NAME, 'INFO',
-             'Attempting to establish initial connection to Elasticsearch host [{}]...'.format(host)),
-            (LOGGER_NAME, 'INFO', '...success!'),
-            (LOGGER_NAME, 'INFO', 'Making sure index [{}] exists...'.format(index)),
-            (LOGGER_NAME, 'INFO', '...index created.')
-        )
+        # Verify the index was created
+        self.assertTrue(self.es.indices.exists(index=index))
 
-    def test_ready_index_exists(self):
-        """ Verify the app does not setup a new Elasticsearch index if one exists already. """
-        host = settings.ELASTICSEARCH['host']
+    def test_alias_exists(self):
+        """ Verify the app does not setup a new Elasticsearch index if the alias is already set. """
         index = settings.ELASTICSEARCH['index']
 
         # Verify the index exists
         self.assertTrue(self.es.indices.exists(index=index))
 
-        with mock.patch.object(IndicesClient, 'create') as mock_create:
-            mock_create.side_effect = TransportError(400)
-
-            with LogCapture(LOGGER_NAME) as l:
-                # This call should NOT raise an exception.
-                call_command('install_es_indexes')
+        call_command('install_es_indexes')
 
         # Verify the index still exists
         self.assertTrue(self.es.indices.exists(index=index))
-
-        l.check(
-            (LOGGER_NAME, 'INFO',
-             'Attempting to establish initial connection to Elasticsearch host [{}]...'.format(host)),
-            (LOGGER_NAME, 'INFO', '...success!'),
-            (LOGGER_NAME, 'INFO', 'Making sure index [{}] exists...'.format(index)),
-            (LOGGER_NAME, 'INFO', '...index already exists.')
-        )
-
-    def test_ready_es_failure(self):
-        """ Verify Elasticsearch errors are raised if the app fails to create the index. """
-        with mock.patch.object(IndicesClient, 'create') as mock_create:
-            mock_create.side_effect = TransportError(500)
-
-            with self.assertRaises(TransportError):
-                call_command('install_es_indexes')

course_discovery/apps/courses/utils.py

+import datetime
+import logging
+
+from course_discovery.apps.courses.config import COURSES_INDEX_CONFIG
+
+logger = logging.getLogger(__name__)
+
+
+class ElasticsearchUtils(object):
+    @classmethod
+    def create_alias_and_index(cls, es, alias):
+        logger.info('Making sure alias [%s] exists...', alias)
+
+        if es.indices.exists_alias(name=alias):
+            # If the alias exists, and points to an open index, we are all set.
+            logger.info('...alias exists.')
+        else:
+            # Create an index with a unique (timestamped) name
+            timestamp = datetime.datetime.utcnow().strftime("%Y%m%d%H%M%S")
+            index = '{alias}_{timestamp}'.format(alias=alias, timestamp=timestamp)
+            es.indices.create(index=index, body=COURSES_INDEX_CONFIG)
+            logger.info('...index [%s] created.', index)
+
+            # Point the alias to the new index
+            body = {
+                'actions': [
+                    {'remove': {'alias': alias, 'index': '*'}},
+                    {'add': {'alias': alias, 'index': index}},
+                ]
+            }
+            es.indices.update_aliases(body)
+            logger.info('...alias updated.')

docs/elasticsearch.rst

+Elasticsearch
+=============
+
+This service uses Elasticsearch to power the course catalog functionality. This allows users to search courses by
+various criteria related to data collected from the E-Commerce Service (Otto) and other potential data sources.
+
+The service is configured to use the `course_discovery` index by default. If you'd like to change the index, or the
+URL of the Elasticsearch service, update the `ELASTICSEARCH` setting.
+
+Elasticsearch has a feature called `aliases`<https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-aliases.html>.
+This feature allows indices to be referenced by an alias. For example, the timestamped course_discovery_20160101113005
+index could be assigned the alias course_discovery. It is common practice to reference aliases in code, rather than
+indices, to allow for index swapping, reindex, and other maintenance without affecting service uptime. We recommend
+following this practice.
+
+Creating an index and alias
+---------------------------
+
+The `install_es_indexes` management command should be used when initializing a new alias and index. This command will
+check to see if the alias exists, and is linked to an open index. If that check is true, the command will exit
+successfully. If that check fails, a new index with a timestamped name (e.g. course_discovery_20160101113005) will be
+created; and, the alias will be assigned to the new index.
+
+.. code-block:: bash
+
+    $ ./manage.py install_es_indexes

docs/index.rst

@@ -4,13 +4,14 @@
    contain the root `toctree` directive.
 
 Course Discovery Service
-===============================
+========================
 A service for serving course discovery and marketing information to partners, mobile, and edX.org
 
 .. toctree::
    :maxdepth: 2
 
    getting_started
+   elasticsearch
    testing
    features
    internationalization