Add optional configuration values for scheduler

CHANGELOG.md

@@ -15,7 +15,7 @@
 - None
 
 ## New features
-- None
+- Expose rule scheduler properties as configurable settings - [#192](https://github.com/jertel/elastalert2/pull/192) - #jertel
 
 ## Other changes
 - Speed up unit tests by adding default parallelism - [164](https://github.com/jertel/elastalert2/pull/164) - @ferozsalam

docs/source/elastalert.rst

@@ -164,6 +164,8 @@ it ran the query for a given rule, and periodically query from that time until t
 this field is a nested unit of time, such as ``minutes: 5``. This is how time is defined in every ElastAlert
 configuration.
 
+``misfire_grace_time``: If the rule scheduler is running behind, due to large numbers of rules or long-running rules, this grace time settings allows a rule to still be executed, provided its next scheduled runt time is no more than this grace period, in seconds, overdue. The default is 5 seconds.
+
 ``writeback_index``: The index on ``es_host`` to use.
 
 ``max_query_size``: The maximum number of documents that will be downloaded from Elasticsearch in a single query. The
@@ -174,6 +176,8 @@ using the size of ``max_query_size`` through the set amount of pages, when ``max
 ``max_scrolling_count``: The maximum amount of pages to scroll through. The default is ``0``, which means the scrolling has no limit.
 For example if this value is set to ``5`` and the ``max_query_size`` is set to ``10000`` then ``50000`` documents will be downloaded at most.
 
+``max_threads``: The maximum number of concurrent threads available to process scheduled rules. Large numbers of long-running rules may require this value be increased, though this could overload the Elasticsearch cluster if too many complex queries are running concurrently. Default is 10.
+
 ``scroll_keepalive``: The maximum time (formatted in `Time Units <https://www.elastic.co/guide/en/elasticsearch/reference/current/common-options.html#time-units>`_) the scrolling context should be kept alive. Avoid using high values as it abuses resources in Elasticsearch, but be mindful to allow sufficient time to finish processing all the results.
 
 ``max_aggregation``: The maximum number of alerts to aggregate together. If a rule has ``aggregation`` set, all

elastalert/elastalert.py

@@ -22,6 +22,7 @@
 import dateutil.tz
 import pytz
 from apscheduler.schedulers.background import BackgroundScheduler
+from apscheduler.executors.pool import ThreadPoolExecutor
 from croniter import croniter
 from elasticsearch.exceptions import ConnectionError
 from elasticsearch.exceptions import ElasticsearchException
@@ -171,7 +172,15 @@ def __init__(self, args):
         self.thread_data.alerts_sent = 0
         self.thread_data.num_hits = 0
         self.thread_data.num_dupes = 0
-        self.scheduler = BackgroundScheduler()
+        executors = {
+            'default': ThreadPoolExecutor(max_workers=self.conf.get('max_threads', 10)),
+        }
+        job_defaults = {
+            'misfire_grace_time': self.conf.get('misfire_grace_time', 5),
+            'coalesce': True,
+            'max_instances': 1
+        }
+        self.scheduler = BackgroundScheduler(executors=executors, job_defaults=job_defaults)
         self.string_multi_field_name = self.conf.get('string_multi_field_name', False)
         self.statsd_instance_tag = self.conf.get('statsd_instance_tag', '')
         self.statsd_host = self.conf.get('statsd_host', '')

elastalert/schema.yaml

@@ -198,6 +198,8 @@ properties:
   query_delay: *timeframe
   max_query_size: {type: integer}
   max_scrolling: {type: integer}
+  max_threads: {type: integer}
+  misfire_grace_time: {type: integer}
 
   owner: {type: string}
   priority: {type: integer}