Merge pull request #458 from RUGSoftEng/backend/uptime_checking

Backend/uptime checking

pydash/pydash_app/__init__.py

@@ -10,6 +10,7 @@
 
 import pydash_app.dashboard.services.fetching
 import pydash_app.dashboard.services.seeding
+import pydash_app.dashboard.services.pinging
 import pydash_app.dashboard
 
 
@@ -26,10 +27,15 @@ def stop_task_scheduler():
 def schedule_periodic_tasks():
     """Schedules all periodic tasks using the default task scheduler, which is declared in pydash.periodic_tasks."""
     import datetime  # <- remove this line when custom interval no longer necessary for testing.
-    dashboard.services.fetching.schedule_all_periodic_dashboards_tasks(
+
+    pydash_app.dashboard.services.fetching.schedule_all_periodic_dashboards_tasks(
         interval=datetime.timedelta(minutes=1)
     )
 
+    # pydash_app.dashboard.services.pinging.schedule_all_periodic_dashboard_pinging(
+    #     interval=datetime.timedelta(seconds=15)
+    # )
+
     pydash_app.user.services.pruning.schedule_periodic_pruning_task()
 
 

pydash/pydash_app/dashboard/downtime.py

@@ -0,0 +1,129 @@
+"""
+Exposes the class DowntimeLog, which keeps track of a web service's downtime,
+and calculates downtime intervals, total downtime, and downtime percentage
+in an on-line manner.
+"""
+
+import persistent
+import datetime
+from collections import defaultdict
+
+
+class DowntimeLog(persistent.Persistent):
+    """
+    Keeps track of downtime, and calculates downtime intervals, total downtime, and downtime percentage
+    in an on-line manner.
+    """
+
+    def __init__(self):
+        self._downtime_intervals = defaultdict(list)  # datetime.date -> list[(datetime.time, datetime.time)]
+        self._total_downtime = defaultdict(datetime.timedelta)  # datetime.date -> datetime.timedelta
+
+        self._downtime_start = None
+
+    def add_ping_result(self, is_up, ping_datetime=datetime.datetime.now(tz=datetime.timezone.utc)):
+        """
+        Add the result of a ping request to the downtime log.
+        :param is_up: Whether the web service is up or not.
+        :param ping_datetime: When the ping took place (approximately); defaults to the current time in UTC.
+        """
+        if is_up:
+            if self._downtime_start:
+                # Split the downtime into intervals of at most 24 hours
+
+                start = self._downtime_start
+                end = min(ping_datetime, _day_end(start))
+
+                while start <= ping_datetime:
+                    date = start.date()
+                    interval = (start.timetz(), end.timetz())
+                    self._downtime_intervals[date].append(interval)
+                    self._total_downtime[date] += (end - start) + datetime.timedelta(microseconds=1)
+
+                    start = _day_start(start + datetime.timedelta(days=1))
+                    end = min(ping_datetime, _day_end(start))
+
+                self._downtime_start = None
+        else:
+            if self._downtime_start is None:
+                self._downtime_start = ping_datetime
+
+    def get_downtime_intervals(
+            self,
+            start=datetime.datetime.now(tz=datetime.timezone.utc).date() - datetime.timedelta(days=90),
+            end=datetime.datetime.now(tz=datetime.timezone.utc).date()):
+        """
+        Return the intervals of downtime per day between two dates.
+        :param start: The start date (exclusive; defaults to 90 days before the current day).
+        :param end: The end date (inclusive; defaults to the current day).
+        :return: A dict containing a list of downtime intervals per day.
+        """
+        if end <= start:
+            ValueError('Date range cannot be negative or zero')
+
+        return {
+            date.strftime('%Y-%m-%d'): list(self._downtime_intervals[date])
+            for date in _date_range(start, end)
+        }
+
+    def get_total_downtimes(
+            self,
+            start=datetime.datetime.now(tz=datetime.timezone.utc).date() - datetime.timedelta(days=90),
+            end=datetime.datetime.now(tz=datetime.timezone.utc).date()):
+        """
+        Return the total amounts of downtime per day between two dates.
+        :param start: The start date (exclusive; defaults to 90 days before the current day).
+        :param end: The end date (inclusive; defaults to the current day).
+        :return: A dict containing the total downtime per day.
+        """
+        if end <= start:
+            raise ValueError('Date range cannot be negative or zero')
+
+        return {
+            date.strftime('%Y-%m-%d'): self._total_downtime[date]
+            for date in _date_range(start, end)
+        }
+
+    def get_downtime_percentage(
+            self,
+            start=datetime.datetime.now(tz=datetime.timezone.utc).date() - datetime.timedelta(days=90),
+            end=datetime.datetime.now(tz=datetime.timezone.utc).date()):
+        """
+        Get the percentage of downtime between two dates.
+        :param start: The start date (exclusive; defaults to 90 days before the current day).
+        :param end: The end date (inclusive; defaults to the current day).
+        :return: A float, the downtime percentage for the given date range.
+        """
+        if end <= start:
+            raise ValueError('Date range cannot be negative or zero')
+
+        total_downtime = sum(
+            (self._total_downtime[date] for date in _date_range(start, end)),
+            datetime.timedelta(0)
+        )
+
+        total_time = end - start
+
+        percentage = total_downtime/total_time*100
+
+        return percentage
+
+
+def _day_start(dt):
+    return datetime.datetime.combine(dt.date(), datetime.time.min).replace(tzinfo=datetime.timezone.utc)
+
+
+def _day_end(dt):
+    return datetime.datetime.combine(dt.date(), datetime.time.max).replace(tzinfo=datetime.timezone.utc)
+
+
+def _date_range(start, end):
+    """
+    Yield dates in the range (start, end].
+    :param start: Start date (exclusive).
+    :param end: End date.
+    """
+    start += datetime.timedelta(days=1)
+    while start <= end:
+        yield start
+        start += datetime.timedelta(days=1)

pydash/pydash_app/dashboard/entity.py

@@ -36,12 +36,14 @@
 import uuid
 import persistent
 from enum import Enum
+from datetime import datetime, timedelta, timezone
+
+from pydash_app.dashboard.downtime import DowntimeLog
 
 from pydash_app.dashboard.endpoint import Endpoint
 from ..dashboard.aggregator import Aggregator
 from pydash_app.dashboard.aggregator.aggregator_group import AggregatorGroup, truncate_datetime_by_granularity
 
-
 class DashboardState(Enum):
     """
     The DashboardState enum indicates the state in which a Dashboard can remain, regarding remote fetching:
@@ -85,13 +87,16 @@ class Dashboard(persistent.Persistent):
     This task is handled by the `dashboard_repository`.
     """
 
-    def __init__(self, url, token, user_id, name=None):
+    def __init__(self, url, token, user_id, name=None, monitor_downtime=False):
         if not isinstance(url, str) or not isinstance(token, str):
             raise TypeError("Dashboard expects both url and token to be strings.")
 
         if name is not None and not isinstance(name, str):
             raise TypeError("Dashboard expects name to be a string.")
 
+        if not isinstance(monitor_downtime, bool):
+            raise TypeError("Dashboard expects monitor_downtime to be a string.")
+
         # Make sure integers and strings are allowed as well.
         if not isinstance(user_id, uuid.UUID):
             user_id = uuid.UUID(user_id)
@@ -111,6 +116,9 @@ def __init__(self, url, token, user_id, name=None):
         self._endpoint_calls = []  # list of unfiltered endpoint calls, for use with an aggregator.
         self._aggregator_group = AggregatorGroup()
 
+        self.monitor_downtime = monitor_downtime
+        self._downtime_log = DowntimeLog()
+
     def __repr__(self):
         return f'<{self.__class__.__name__} id={self.id} url={self.url}>'
 
@@ -262,7 +270,7 @@ def statistic(self, statistic, filters={}):
         :raises KeyError: This happens when the statistic is not supported by the dashboard.
         """
         return self._aggregator_group.fetch_aggregator(filters).as_dict()[statistic]
-
+    
     def statistic_per_timeslice(self, statistic, timeslice, timeslice_is_static, start_datetime, end_datetime, filters={}):
         """
         Slices up the specified datetime range (=[start_datetime, end_datetime)) into slices of the size of `timeslice`.
@@ -334,3 +342,33 @@ def statistic_per_timeslice(self, statistic, timeslice, timeslice_is_static, sta
             return_dict[datetime] = aggregator.as_dict()[statistic]
 
         return return_dict
+
+    def add_ping_result(self, is_up, ping_datetime=datetime.now(tz=timezone.utc)):
+        """
+        Adds the result of a ping request to the dashboard.
+        :param is_up: Whether the dashboard's web service is up.
+        :param ping_datetime: When the ping took place (approximately); defaults to the current time in UTC.
+        """
+        self._downtime_log.add_ping_result(is_up, ping_datetime)
+
+    def get_downtime_data(
+            self,
+            start=datetime.now(tz=timezone.utc).date() - timedelta(days=90),
+            end=datetime.now(tz=timezone.utc).date()):
+        """
+        Returns a dict containing this dashboard's downtime data for a given date range.
+        :param start: The start date (exclusive; defaults to 90 days before the current date).
+        :param end: The end date (inclusive; defaults to the current date).
+        :return: A dictionary containing the dashboard's downtime data in the given date range.
+        """
+        return {
+            'downtime_intervals': self._downtime_log.get_downtime_intervals(start, end),
+            'total_downtimes': self._downtime_log.get_total_downtimes(start, end),
+            'downtime_percentage': self._downtime_log.get_downtime_percentage(start, end)
+        }
+
+    # Required because `multi_indexed_collection` puts dashboards in a set,
+    #  that needs to order its keys for fast lookup.
+    # Because the IDs are unchanging integer values, use that.
+    def __lt__(self, other):
+        return self.id < other.id

pydash/pydash_app/dashboard/services/fetching.py

@@ -40,17 +40,17 @@ def schedule_all_periodic_dashboards_tasks(
 
 
 def schedule_periodic_dashboard_fetching(
-        dashboard,
+        dashboard_id,
         interval=timedelta(hours=1),
         scheduler=periodic_tasks.default_task_scheduler):
     """
     Schedules the periodic EndpointCall fetching task for this dashboard.
     """
-    logger.info(f'Creating periodic fetching task for {dashboard}')
+    logger.info(f'Creating periodic fetching task for {dashboard_id}')
 
     periodic_tasks.add_periodic_task(
-        name=("dashboard", dashboard.id, "fetching"),
-        task=partial(fetch_and_update_new_dashboard_info, dashboard.id),
+        name=("dashboard", dashboard_id, "fetching"),
+        task=partial(fetch_and_update_new_dashboard_info, dashboard_id),
         interval=interval,
         scheduler=scheduler)
 
@@ -62,16 +62,17 @@ def schedule_historic_dashboard_fetching(
     The periodic fetching of new EndpointCall information is scheduled as soon as this task completes.
     """
 
-    def task(dashboard_id):
-        fetch_and_update_historic_dashboard_info(dashboard_id)
-        schedule_periodic_dashboard_fetching(dashboard_id)
-
     periodic_tasks.add_background_task(
         name=("dashboard", dashboard.id, "historic_fetching"),
-        task=partial(task, dashboard.id),
+        task=partial(_task_historic_dashboard_fetching, dashboard.id),
         scheduler=scheduler)
 
 
+def _task_historic_dashboard_fetching(dashboard_id):
+    fetch_and_update_historic_dashboard_info(dashboard_id)
+    schedule_periodic_dashboard_fetching(dashboard_id)
+
+
 def fetch_and_update_new_dashboard_info(dashboard_id):
     """
     Updates the dashboard with the new EndpointCall information that is fetched from the Dashboard's remote location.

pydash/pydash_app/dashboard/services/pinging.py

@@ -0,0 +1,78 @@
+"""
+Periodically pings a dashboard to see if the web service is still up.
+"""
+
+from functools import partial
+from datetime import timedelta
+import json
+
+import requests
+
+import flask_monitoring_dashboard_client
+import pydash_app.dashboard.repository as dashboard_repository
+import pydash_logger
+import periodic_tasks
+
+logger = pydash_logger.Logger(__name__)
+
+_DEFAULT_PING_INTERVAL = timedelta(minutes=5)
+
+
+def schedule_all_periodic_dashboard_pinging(
+        interval=_DEFAULT_PING_INTERVAL,
+        scheduler=periodic_tasks.default_task_scheduler):
+    """
+    Set up periodic dashboard pinging tasks for all dashboards that want their uptime to be monitored.
+    :param interval: The frequency with which to ping a dashboard, defaults to 5 minutes.
+    :param scheduler: The task scheduler to schedule the tasks to, defaults to the default scheduler.
+    """
+    for dashboard in dashboard_repository.all():
+        schedule_periodic_dashboard_pinging(dashboard, interval, scheduler)
+
+
+def schedule_periodic_dashboard_pinging(
+        dashboard,
+        interval=_DEFAULT_PING_INTERVAL,
+        scheduler=periodic_tasks.default_task_scheduler):
+    """
+    Set up a periodic pinging task for a dashboard if the dashboard allows it.
+    :param dashboard: The dashboard to set up a pinging task for.
+    :param interval: The frequency with which to ping a dashboard, defaults to 5 minutes.
+    :param scheduler: The task scheduler to schedule this task to, defaults to the default scheduler.
+    """
+
+    if dashboard.monitor_downtime:
+        periodic_tasks.add_periodic_task(
+            name=('dashboard', dashboard.id, 'pinging'),
+            task=partial(_ping_dashboard, dashboard.id),
+            interval=interval,
+            scheduler=scheduler)
+
+
+def _ping_dashboard(dashboard_id):
+    try:
+        dashboard = dashboard_repository.find(dashboard_id)
+    except KeyError:
+        logger.warning('Dashboard does not exist')
+        return
+
+    is_up = _is_dashboard_up(dashboard.url)
+    dashboard.add_ping_result(is_up)
+
+    dashboard_repository.update(dashboard)
+
+
+def _is_dashboard_up(url):
+    """
+    Connect to a dashboard to see if it's up.
+    :param url: The dashboard's URL.
+    :return: True or False depending on whether the dashboard is up.
+    """
+    try:
+        flask_monitoring_dashboard_client.get_details(url)
+    except requests.exceptions.RequestException:
+        return False
+    except (json.JSONDecodeError, Exception):
+        return True
+
+    return True

pydash/pydash_app/dashboard/services/seeding.py

@@ -25,7 +25,8 @@ def seed():
         dashboard_new = Dashboard('http://flask-sample.koenbolhuis.nl/dashboard',
                                   'cc83733cb0af8b884ff6577086b87909',
                                   user.get_id(),
-                                  'Testing Dashboard (FMD v1.12.0)')
+                                  'Testing Dashboard (FMD v1.12.0+)',
+                                  True)
         dashboard_old = Dashboard('http://flask-sample-old.koenbolhuis.nl/dashboard',
                                   'cc83733cb0af8b884ff6577086b87909',
                                   user.get_id(),

pydash/pydash_app/user/services/pruning.py

@@ -20,7 +20,7 @@ def schedule_periodic_pruning_task(
     :param scheduler: The TaskScheduler instance that should schedule this user pruning task and execute it.
       Defaults to the default task scheduler of pydash.periodic_tasks.
     """
-    scheduler.add_periodic_task(
+    periodic_tasks.add_periodic_task(
         name=('users', 'pruning'),
         task=_prune_unverified_users,
         interval=interval,