argos server cleandb now keep a number of results per task

README.md

@@ -1,15 +1,27 @@
 # Argos monitoring
 
-Argos is an HTTP monitoring service. It allows you to define a list of websites to monitor, and a list of checks to run on these websites. It will then run these checks periodically, and alert you if something goes wrong.
+An HTTP monitoring service.
 
-See the online documentation at http://framasoft.frama.io/framaspace/argos
+1. Define a list of websites to monitor
+2. Specify a list of checks to run on these websites.
+3. Argos will run the checks periodically and alert you if something goes wrong.
 
-Todo:
+Internally, a HTTP API is exposed, and a job queue is used to distribute the checks to the agents.
+
+- [Online documentation](http://framasoft.frama.io/framaspace/argos)
+- [Issue tracker](https://framagit.org/framasoft/framaspace/argos/-/issues)
+
+## Resquirements
+
+- **Python**: 3.11+
+- **Backends**: SQLite (development), PostgreSQL 14+ (production)
+
+## Todo:
 
-- [ ] Cleandb should keep max number of results by task
 - [ ] Do not return empty list on / when no results from agents.
 - [ ] Last seen agents
 - [ ] donner un aperçu rapide de l’état de la supervision.
+- [ ] Allow passing a dict to check
 - [ ] Rename error in unexpected error
 - [ ] Use background tasks for alerting
 - [ ] Delete outdated tasks from config

argos/commands.py

@@ -64,13 +64,13 @@ def start(host, port, reload, log_config):
 
 
 @server.command()
-@click.option("--max-results", default=100, help="Maximum number of results to keep")
+@click.option("--max-results", default=100, help="Number of results per tasks to keep")
 @click.option(
     "--max-lock-seconds",
     default=100,
     help="The number of seconds after which a lock is considered stale",
 )
-def clean(max_results, max_lock_seconds):
+def cleandb(max_results, max_lock_seconds):
     """Clean the database (to run routinely)
 
     - Removes old results from the database.

argos/server/queries.py

@@ -117,16 +117,27 @@ async def get_severity_counts(db: Session):
 
 
 async def remove_old_results(db: Session, max_results: int):
-    # Get the id of the oldest result to keep, then delete all results older than that
-    subquery = (
-        db.query(Result.id).order_by(desc(Result.id)).limit(max_results).subquery()
-    )
-    min_id = db.query(func.min(subquery.c.id)).scalar()
-    if min_id:
-        deleted = db.query(Result).where(Result.id < min_id).delete()
-        db.commit()
-    else:
-        deleted = 0
+    tasks = db.query(Task).all()
+    deleted = 0
+    for task in tasks:
+        # Get the id of the oldest result to keep
+        subquery = (
+            db.query(Result.id)
+            .filter(Result.task_id == task.id)
+            .order_by(desc(Result.id))
+            .limit(max_results)
+            .subquery()
+        )
+        min_id = db.query(func.min(subquery.c.id)).scalar()
+
+        # Delete all the results older than min_id
+        if min_id:
+            deleted += (
+                db.query(Result)
+                .where(Result.id < min_id, Result.task_id == task.id)
+                .delete()
+            )
+            db.commit()
 
     return deleted
 

docs/cli.md

@@ -102,7 +102,7 @@ Options:
 ### Server clean
 <!--
 .. [[[cog
-    help(["server", "clean", "--help"])
+    help(["server", "cleandb", "--help"])
 .. ]]] -->
 
 ```man

tests/test_queries.py

@@ -7,13 +7,28 @@ from argos.server.models import Result, Task
 
 
 @pytest.mark.asyncio
-async def test_remove_old_results(db, ten_results):
-    assert db.query(Result).count() == 10
+async def test_remove_old_results(db, ten_tasks):
+    for task in ten_tasks:
+        for i in range(5):
+            result = Result(
+                submitted_at=datetime.now(),
+                status="success",
+                context={"foo": "bar"},
+                task=task,
+                agent_id="test",
+                severity="ok",
+            )
+            db.add(result)
+    db.commit()
+
+    # So we have 5 results per tasks
+    assert db.query(Result).count() == 50
+    # Keep only 2
     deleted = await queries.remove_old_results(db, 2)
-    assert deleted == 8
-    assert db.query(Result).count() == 2
-    # We should keep the last two results
-    assert db.query(Result).all() == ten_results[-2:]
+    assert deleted == 30
+    assert db.query(Result).count() == 20
+    for task in ten_tasks:
+        assert db.query(Result).filter(Result.task == task).count() == 2
 
 
 @pytest.mark.asyncio