Announce maintenance in stable deployments (#5979)

lambdas/service/app.py

@@ -73,6 +73,9 @@
 from azul.logging import (
     configure_app_logging,
 )
+from azul.maintenance import (
+    MaintenanceService,
+)
 from azul.openapi import (
     application_json,
     format_description as fd,
@@ -232,7 +235,7 @@
         # changes and reset the minor version to zero. Otherwise, increment only
         # the minor version for backwards compatible changes. A backwards
         # compatible change is one that does not require updates to clients.
-        'version': '9.1'
+        'version': '9.2'
     },
     'tags': [
         {
@@ -870,6 +873,44 @@ def get_integrations():
                     body=json.dumps(body))
 
 
+@app.route(
+    '/maintenance/schedule',
+    methods=['GET'],
+    cors=True,
+    method_spec={
+        'summary': 'A maintenance schedule as an JSON object',
+        'tags': ['Auxiliary'],
+        'responses': {
+            '200': {
+                'description': fd('''
+                    This object may be hanceforth refered to as "the schedule"
+                    or `schedule`.
+                    The `start` time of an event is its `actual_start` if set,
+                    or its `planned_start` otherwise. The `end` time of an event
+                    is its `actual_end` if set, or its `start` plus
+                    `planned_duration` otherwise. All events in the schedule are
+                    sorted by their `start` time. No two events have the same
+                    `start` time. Each event defines an interval
+                    `[e.start, e.end)` and there is no overlap between these
+                     intervals.
+
+                     A pending event is one where `actual_start` is absent. An
+                     active event is one where `actual_start` is present but
+                     `actual_end` is absent. There can be at most one active
+                     event.
+                ''')
+            }
+        }
+    }
+)
+def get_maintenance_schedule():
+    service = MaintenanceService()
+    schedule = service.get_schedule.to_json()
+    return Response(status_code=200,
+                    headers={'content-type': 'application/json'},
+                    body=json.dumps(schedule))
+
+
 @app.route(
     '/index/catalogs',
     methods=['GET'],

lambdas/service/openapi.json

@@ -3,7 +3,7 @@
     "info": {
         "title": "azul_service",
         "description": "\n# Overview\n\nAzul is a REST web service for querying metadata associated with\nboth experimental and analysis data from a data repository. In order\nto deliver response times that make it suitable for interactive use\ncases, the set of metadata properties that it exposes for sorting,\nfiltering, and aggregation is limited. Azul provides a uniform view\nof the metadata over a range of diverse schemas, effectively\nshielding clients from changes in the schemas as they occur over\ntime. It does so, however, at the expense of detail in the set of\nmetadata properties it exposes and in the accuracy with which it\naggregates them.\n\nAzul denormalizes and aggregates metadata into several different\nindices for selected entity types. Metadata entities can be queried\nusing the [Index](#operations-tag-Index) endpoints.\n\nA set of indices forms a catalog. There is a default catalog called\n`dcp2` which will be used unless a\ndifferent catalog name is specified using the `catalog` query\nparameter. Metadata from different catalogs is completely\nindependent: a response obtained by querying one catalog does not\nnecessarily correlate to a response obtained by querying another\none. Two catalogs can contain metadata from the same sources or\ndifferent sources. It is only guaranteed that the body of a\nresponse by any given endpoint adheres to one schema,\nindependently of which catalog was specified in the request.\n\nAzul provides the ability to download data and metadata via the\n[Manifests](#operations-tag-Manifests) endpoints. The\n`curl` format manifests can be used to\ndownload data files. Other formats provide various views of the\nmetadata. Manifests can be generated for a selection of files using\nfilters. These filters are interchangeable with the filters used by\nthe [Index](#operations-tag-Index) endpoints.\n\nAzul also provides a [summary](#operations-Index-get_index_summary)\nview of indexed data.\n\n## Data model\n\nAny index, when queried, returns a JSON array of hits. Each hit\nrepresents a metadata entity. Nested in each hit is a summary of the\nproperties of entities associated with the hit. An entity is\nassociated either by a direct edge in the original metadata graph,\nor indirectly as a series of edges. The nested properties are\ngrouped by the type of the associated entity. The properties of all\ndata files associated with a particular sample, for example, are\nlisted under `hits[*].files` in a `/index/samples` response. It is\nimportant to note that while each _hit_ represents a discrete\nentity, the properties nested within that hit are the result of an\naggregation over potentially many associated entities.\n\nTo illustrate this, consider a data file that is part of two\nprojects (a project is a group of related experiments, typically by\none laboratory, institution or consortium). Querying the `files`\nindex for this file yields a hit looking something like:\n\n```\n{\n    \"projects\": [\n        {\n            \"projectTitle\": \"Project One\"\n            \"laboratory\": ...,\n            ...\n        },\n        {\n            \"projectTitle\": \"Project Two\"\n            \"laboratory\": ...,\n            ...\n        }\n    ],\n    \"files\": [\n        {\n            \"format\": \"pdf\",\n            \"name\": \"Team description.pdf\",\n            ...\n        }\n    ]\n}\n```\n\nThis example hit contains two kinds of nested entities (a hit in an\nactual Azul response will contain more): There are the two projects\nentities, and the file itself. These nested entities contain\nselected metadata properties extracted in a consistent way. This\nmakes filtering and sorting simple.\n\nAlso notice that there is only one file. When querying a particular\nindex, the corresponding entity will always be a singleton like\nthis.\n",
-        "version": "9.1"
+        "version": "9.2"
     },
     "tags": [
         {
@@ -653,6 +653,19 @@
                 }
             }
         },
+        "/maintenance/schedule": {
+            "get": {
+                "summary": "A maintenance schedule as an JSON object",
+                "tags": [
+                    "Auxiliary"
+                ],
+                "responses": {
+                    "200": {
+                        "description": "\nThis object may be hanceforth refered to as \"the schedule\"\nor `schedule`.\nThe `start` time of an event is its `actual_start` if set,\nor its `planned_start` otherwise. The `end` time of an event\nis its `actual_end` if set, or its `start` plus\n`planned_duration` otherwise. All events in the schedule are\nsorted by their `start` time. No two events have the same\n`start` time. Each event defines an interval\n`[e.start, e.end)` and there is no overlap between these\n intervals.\n\n A pending event is one where `actual_start` is absent. An\n active event is one where `actual_start` is present but\n `actual_end` is absent. There can be at most one active\n event.\n"
+                    }
+                }
+            }
+        },
         "/index/catalogs": {
             "get": {
                 "summary": "List all available catalogs.",

scripts/manage_maintenance.py

@@ -0,0 +1,142 @@
+"""
+This is a command line utility for managing announcement of maintenance events.
+Reads the JSON from designated bucket, deserializes the model from it, validates
+the model, applies an action to it, serializes the model back to JSON and
+finally uploads it back to the bucket where the service exposes it. The service
+must also validate the model before returning it.
+"""
+import argparse
+from datetime import (
+    timedelta,
+)
+import json
+import re
+import sys
+
+from azul import (
+    require,
+)
+from azul.args import (
+    AzulArgumentHelpFormatter,
+)
+from azul.maintenance import (
+    MaintenanceService,
+)
+
+
+def parse_duration(duration: str) -> timedelta:
+    """
+    >>> parse_duration('1d')
+    datetime.timedelta(days=1)
+    >>> parse_duration('24 hours')
+    datetime.timedelta(days=1)
+    >>> parse_duration('.5 Days 12 hours')
+    datetime.timedelta(days=1)
+
+    >>> parse_duration('2h20Min')
+    datetime.timedelta(seconds=8400)
+    >>> parse_duration('1 H 80m')
+    datetime.timedelta(seconds=8400)
+    >>> parse_duration('140 Minutes')
+    datetime.timedelta(seconds=8400)
+
+    >>> parse_duration('2 Days 3hours 4min 5 secs')
+    datetime.timedelta(days=2, seconds=11045)
+    >>> parse_duration('1d 25h')
+    datetime.timedelta(days=2, seconds=3600)
+    >>> parse_duration('1m30s')
+    datetime.timedelta(seconds=90)
+
+    >>> parse_duration('Bad foo')
+    Traceback (most recent call last):
+    ...
+    azul.RequirementError: Try a duration such as "2d 6hrs", "1.5 Days", or "15m"
+    """
+
+    pattern = r'(\d*\.?\d+)\s*(d|h|m|s)'
+    matches = re.findall(pattern, duration.lower())
+    require(bool(matches), 'Try a duration such as "2d 6hrs", "1.5 Days", or "15m"')
+    time_delta = {'days': 0, 'hours': 0, 'minutes': 0, 'seconds': 0}
+    for value, unit in matches:
+        value = float(value)
+        match unit:
+            case 'd':
+                time_delta['days'] += value
+            case 'h':
+                time_delta['hours'] += value
+            case 'm':
+                time_delta['minutes'] += value
+            case 's':
+                time_delta['seconds'] += value
+    return timedelta(**time_delta)
+
+
+def main(args: list[str]):
+    parser = argparse.ArgumentParser(description=__doc__,
+                                     formatter_class=AzulArgumentHelpFormatter)
+    subparsers = parser.add_subparsers(dest='command')
+    list_parser = subparsers.add_parser('list', help='List events in JSON form')
+    list_parser.add_argument('--all', action='store_true',
+                             help='Include completed events')
+    add_parser = subparsers.add_parser('add', help='Schedule an event')
+    add_parser.add_argument('--start', required=True,
+                            help='Event start time (ISO format)')
+    add_parser.add_argument('--duration', required=True,
+                            help='Event duration (e.g., "1h30m", "2d")')
+    add_parser.add_argument('--description', required=True,
+                            help='Event description')
+    add_parser.add_argument('--partial-responses', nargs='+',
+                            help='Catalog names for partial responses')
+    add_parser.add_argument('--degraded-performance', nargs='+',
+                            help='Catalog names for degraded performance')
+    add_parser.add_argument('--service-unavailable', nargs='+',
+                            help='Catalog names for service unavailability')
+    cancel_parser = subparsers.add_parser('cancel',
+                                          help='Cancel a pending event')
+    cancel_parser.add_argument('--index', type=int, required=True,
+                               help='Index of the event to cancel')
+    subparsers.add_parser('start', help='Activate a pending event')
+    subparsers.add_parser('end', help='Complete the active event')
+    adjust_parser = subparsers.add_parser('adjust',
+                                          help='Modify the active event')
+    adjust_parser.add_argument('--duration', required=True,
+                               help='New event duration (e.g., "1h30m", "2d")')
+
+    args = parser.parse_args(args)
+
+    service = MaintenanceService()
+
+    if args.command == 'list':
+        events = service.get_schedule
+        if args.all:
+            events = events.to_json()
+        else:
+            active = events.active_event()
+            active = {} if active is None else {'active': active.to_json()}
+            pending = events.pending_events()
+            pending = {'pending': [pe.to_json() for pe in pending]}
+            events = active | pending
+    elif args.command == 'add':
+        duration = int(parse_duration(args.duration).total_seconds())
+        events = service.provision_event(planned_start=args.start,
+                                         planned_duration=duration,
+                                         description=args.description,
+                                         partial=args.partial_responses,
+                                         degraded=args.degraded_performance,
+                                         unavailable=args.service_unavailable)
+        events = service.add_event(events)
+    elif args.command == 'cancel':
+        events = service.cancel_event(args.index)
+    elif args.command == 'start':
+        events = service.start_event()
+    elif args.command == 'end':
+        events = service.end_event()
+    elif args.command == 'adjust':
+        events = service.adjust_event(parse_duration(args.duration))
+    else:
+        raise argparse.ArgumentError(subparsers, 'Invalid command')
+    print(json.dumps(events, indent=4))
+
+
+if __name__ == '__main__':
+    main(sys.argv[1:])