Bug 1660506 - [taskgraph] Ensure backstop determination uses time since last backstop, r=marco

Backstop pushes have a time component so that we can guarantee they run at least every N
hours. Prior, they measured the time since the last push. This means that as long as a new
push comes in at least once every 4 hours, the backstop would keep being pushed out.

This means we could theoretically go 80 hours without a backstop push, which makes finding
merge candidates tricky.

This change ensures we measure the 4 hours as time since last backstop, as opposed to time
since last push.

Differential Revision: https://phabricator.services.mozilla.com/D88913

taskcluster/taskgraph/test/test_util_backstop.py

@@ -5,12 +5,19 @@
 from __future__ import absolute_import, print_function, unicode_literals
 
 from datetime import datetime
+from textwrap import dedent
 from time import mktime
 
 import pytest
 from mozunit import main
 
-from taskgraph.util.backstop import is_backstop
+from taskgraph.util.backstop import (
+    is_backstop,
+    BACKSTOP_INDEX,
+    BACKSTOP_PUSH_INTERVAL,
+    BACKSTOP_TIME_INTERVAL,
+)
+from taskgraph.util.taskcluster import get_index_url
 
 
 @pytest.fixture(scope='module')
@@ -20,50 +27,46 @@ def params():
         'head_repository': 'https://hg.mozilla.org/integration/autoland',
         'head_rev': 'abcdef',
         'project': 'autoland',
-        'pushlog_id': 1,
         'pushdate': mktime(datetime.now().timetuple()),
     }
 
 
 def test_is_backstop(responses, params):
+    url = get_index_url(
+        BACKSTOP_INDEX.format(project=params["project"])
+    ) + "/artifacts/public/parameters.yml"
 
     responses.add(
         responses.GET,
-        "https://hg.mozilla.org/integration/autoland/json-pushes/?version=2&startID=16&endID=17",  # noqa
-        json={"pushes": {"17": {}}},
-        status=200,
+        url,
+        status=404,
     )
 
     # If there's no previous push date, run tasks
-    params['pushlog_id'] = 18
+    params["pushlog_id"] = 1
     assert is_backstop(params)
 
-    responses.add(
+    responses.replace(
         responses.GET,
-        "https://hg.mozilla.org/integration/autoland/json-pushes/?version=2&startID=17&endID=18",  # noqa
-        json={"pushes": {"18": {"date": params['pushdate']}}},
+        url,
+        body=dedent("""
+        pushdate: {pushdate}
+        """.format(pushdate=params["pushdate"])),
         status=200,
     )
 
-    # Only multiples of 20 schedule tasks. Pushdate from push 19 was cached.
-    params['pushlog_id'] = 19
-    params['pushdate'] += 3599
+    # Only multiples of push interval schedule tasks.
+    params['pushlog_id'] = BACKSTOP_PUSH_INTERVAL - 1
+    params['pushdate'] += 1
     assert not is_backstop(params)
 
-    params['pushlog_id'] = 20
+    params['pushlog_id'] = BACKSTOP_PUSH_INTERVAL
     params['pushdate'] += 1
     assert is_backstop(params)
 
-    responses.add(
-        responses.GET,
-        "https://hg.mozilla.org/integration/autoland/json-pushes/?version=2&startID=19&endID=20",  # noqa
-        json={"pushes": {"20": {"date": params['pushdate']}}},
-        status=200,
-    )
-
-    # Tasks are also scheduled if four hours have passed.
-    params['pushlog_id'] = 21
-    params['pushdate'] += 4 * 3600
+    # Tasks are also scheduled if the time interval has passed.
+    params['pushlog_id'] = BACKSTOP_PUSH_INTERVAL + 1
+    params['pushdate'] += BACKSTOP_TIME_INTERVAL * 60
     assert is_backstop(params)
 
 

taskcluster/taskgraph/util/backstop.py

@@ -4,15 +4,14 @@
 
 from __future__ import absolute_import, print_function, unicode_literals
 
-import logging
+from requests import HTTPError
 
-from taskgraph.util.hg import get_push_data
+from taskgraph.util.taskcluster import get_artifact_from_index
 
 
 BACKSTOP_PUSH_INTERVAL = 20
 BACKSTOP_TIME_INTERVAL = 60 * 4  # minutes
-
-logger = logging.getLogger(__name__)
+BACKSTOP_INDEX = "gecko.v2.{project}.latest.taskgraph.backstop"
 
 
 def is_backstop(
@@ -42,31 +41,20 @@ def is_backstop(
     if pushid % push_interval == 0:
         return True
 
-    # We also want to ensure we run all tasks at least once per N minutes.
-    if (
-        time_interval > 0
-        and minutes_between_pushes(
-            time_interval, params["head_repository"], project, pushid, pushdate
-        )
-        >= time_interval
-    ):
-        return True
+    if time_interval <= 0:
         return False
 
+    # We also want to ensure we run all tasks at least once per N minutes.
+    index = BACKSTOP_INDEX.format(project=project)
 
-def minutes_between_pushes(time_interval, repository, project, cur_push_id, cur_push_date):
-    # figure out the minutes that have elapsed between the current push and previous one
-    # defaulting to max min so if we can't get value, defaults to run the task
-    min_between_pushes = time_interval
-    prev_push_id = cur_push_id - 1
+    try:
+        last_pushdate = get_artifact_from_index(index, 'public/parameters.yml')["pushdate"]
+    except HTTPError as e:
+        if e.response.status_code == 404:
+            # There hasn't been a backstop push yet.
+            return True
+        raise
 
-    data = get_push_data(repository, project, prev_push_id, prev_push_id)
-
-    if data is not None:
-        prev_push_date = data[prev_push_id].get('date', 0)
-
-        # now have datetime of current and previous push
-        if cur_push_date > 0 and prev_push_date > 0:
-            min_between_pushes = (cur_push_date - prev_push_date) / 60
-
-    return min_between_pushes
+    if (pushdate - last_pushdate) / 60 >= time_interval:
+        return True
+    return False