cloud run tasks

Author: Strvm

README.md

@@ -133,7 +133,6 @@ Most local contributor workflows only need defaults, but these are commonly used
 - `IS_IT_DOWN_PROXY_SECRET_PROJECT_ID`: GCP project containing checker proxy secrets.
 - `IS_IT_DOWN_DEFAULT_CHECKER_PROXY_SECRET_ID`: default proxy secret ID.
 - `IS_IT_DOWN_CHECKER_CONCURRENCY` (default: `10`)
-- `IS_IT_DOWN_CHECKER_TASK_BATCH_SIZE` (default: `50`; Cloud Run task shard batch size for scheduled checker jobs)
 - `IS_IT_DOWN_CHECKER_MAX_RESPONSE_BODY_BYTES` (default: `524288`)
 - `IS_IT_DOWN_CHECKER_MAX_JSON_RESPONSE_BODY_BYTES` (default: `1048576`)
 - `IS_IT_DOWN_CHECKER_INSERT_BATCH_SIZE` (default: `500`)

infra/terraform/README.md

@@ -46,7 +46,6 @@ File layout:
 Checker job defaults:
 
 - `checker_job_task_count = 4` (Cloud Run Job tasks and parallelism)
-- `checker_task_batch_size = 50` (max checker classes per shard batch in each task)
 
 ## Example: dev apply
 

infra/terraform/cloud_run.tf

@@ -91,11 +91,6 @@ resource "google_cloud_run_v2_job" "checker" {
           value = tostring(var.checker_concurrency)
         }
 
-        env {
-          name  = "IS_IT_DOWN_CHECKER_TASK_BATCH_SIZE"
-          value = tostring(var.checker_task_batch_size)
-        }
-
         env {
           name  = "IS_IT_DOWN_BIGQUERY_PROJECT_ID"
           value = var.project[terraform.workspace]

infra/terraform/variables.tf

@@ -151,11 +151,6 @@ variable "checker_job_task_count" {
   default = 4
 }
 
-variable "checker_task_batch_size" {
-  type    = number
-  default = 50
-}
-
 variable "api_cache_enabled" {
   type    = bool
   default = true

src/is_it_down/checkers/services/npmjs.py

@@ -54,12 +54,12 @@ class NpmjsRegistryWebCheck(HtmlMarkerCheck):
     """Represent `NpmjsRegistryWebCheck`."""
 
     check_key = "npmjs_registryweb"
-    endpoint_key = "https://registry.npmjs.org/react"
+    endpoint_key = "https://registry.npmjs.org/react/latest"
     interval_seconds = 60
     timeout_seconds = 5.0
     required_content_type_fragment = None
     weight = 0.15
-    expected_markers = ("name",)
+    expected_markers = ("name", "version")
 
 
 class NpmjsServiceChecker(BaseServiceChecker):

src/is_it_down/scripts/run_scheduled_checks.py

@@ -123,37 +123,31 @@ def _shard_service_checker_classes(
     *,
     task_index: int,
     task_count: int,
-    checkers_per_task: int,
 ) -> list[type[BaseServiceChecker]]:
     """Shard service checker classes for a Cloud Run Job task.
 
     Args:
         service_checker_classes: All checker classes sorted in deterministic order.
         task_index: The zero-based Cloud Run Job task index.
         task_count: The total number of Cloud Run Job tasks for this execution.
-        checkers_per_task: Maximum number of checkers in each shard batch.
 
     Returns:
         The checker classes assigned to this task.
 
     Raises:
-        RuntimeError: If there are not enough Cloud Run tasks for the configured batch size.
-        ValueError: If `task_count` or `checkers_per_task` is not positive.
+        ValueError: If `task_count` is not positive.
     """
     if task_count <= 0:
         raise ValueError("task_count must be greater than 0.")
-    if checkers_per_task <= 0:
-        raise ValueError("checkers_per_task must be greater than 0.")
 
-    required_task_count = (len(service_checker_classes) + checkers_per_task - 1) // checkers_per_task
-    if task_count < required_task_count:
-        raise RuntimeError(
-            f"Cloud Run task count ({task_count}) is too small for {len(service_checker_classes)} service checkers "
-            f"with checkers_per_task={checkers_per_task}. Increase task_count to at least {required_task_count}."
-        )
+    total_checker_count = len(service_checker_classes)
+    base_shard_size = total_checker_count // task_count
+    remainder = total_checker_count % task_count
 
-    start = task_index * checkers_per_task
-    end = start + checkers_per_task
+    # Distribute the remainder one-by-one to the first tasks for an even split.
+    shard_size = base_shard_size + (1 if task_index < remainder else 0)
+    start = task_index * base_shard_size + min(task_index, remainder)
+    end = start + shard_size
     return list(service_checker_classes[start:end])
 
 
@@ -180,20 +174,16 @@ def _resolve_service_checker_classes(
         return service_checker_classes
 
     task_index, task_count = task_metadata
-    settings = get_settings()
-    checkers_per_task = max(1, settings.checker_task_batch_size)
     sharded_service_checker_classes = _shard_service_checker_classes(
         service_checker_classes,
         task_index=task_index,
         task_count=task_count,
-        checkers_per_task=checkers_per_task,
     )
 
     logger.info(
         "checker_job.task_shard_assigned",
         cloud_run_task_index=task_index,
         cloud_run_task_count=task_count,
-        checkers_per_task=checkers_per_task,
         total_checker_count=len(service_checker_classes),
         assigned_checker_count=len(sharded_service_checker_classes),
     )

src/is_it_down/settings.py

@@ -36,7 +36,6 @@ class Settings(BaseSettings):
     worker_lease_seconds: int = 30
     worker_max_attempts: int = 3
     checker_concurrency: int = 10
-    checker_task_batch_size: int = 50
     checker_max_response_body_bytes: int = 524288
     checker_max_json_response_body_bytes: int = 1048576
     checker_insert_batch_size: int = 500

tests/test_run_scheduled_checks.py

@@ -131,9 +131,6 @@ def test_resolve_cloud_run_task_metadata_returns_none_without_env(monkeypatch: p
 
 
 def test_resolve_service_checker_classes_shards_by_task(monkeypatch: pytest.MonkeyPatch) -> None:
-    monkeypatch.setenv("IS_IT_DOWN_CHECKER_TASK_BATCH_SIZE", "2")
-    _reset_settings_cache()
-
     checkers = {
         key: _build_dummy_checker(key)
         for key in ["alpha", "bravo", "charlie", "delta", "echo", "foxtrot"]
@@ -149,20 +146,22 @@ def test_resolve_service_checker_classes_shards_by_task(monkeypatch: pytest.Monk
     assert [checker.service_key for checker in task_two] == ["echo", "foxtrot"]
 
 
-def test_resolve_service_checker_classes_raises_when_task_count_too_low(
+def test_resolve_service_checker_classes_evenly_distributes_remainder(
     monkeypatch: pytest.MonkeyPatch,
 ) -> None:
-    monkeypatch.setenv("IS_IT_DOWN_CHECKER_TASK_BATCH_SIZE", "2")
-    _reset_settings_cache()
-
     checkers = {
         key: _build_dummy_checker(key)
-        for key in ["alpha", "bravo", "charlie", "delta", "echo", "foxtrot"]
+        for key in ["alpha", "bravo", "charlie", "delta", "echo", "foxtrot", "golf"]
     }
     monkeypatch.setattr(run_scheduled_checks, "discover_service_checkers", lambda: checkers)
 
-    with pytest.raises(RuntimeError, match="too small"):
-        run_scheduled_checks._resolve_service_checker_classes([], task_metadata=(0, 2))
+    task_zero = run_scheduled_checks._resolve_service_checker_classes([], task_metadata=(0, 3))
+    task_one = run_scheduled_checks._resolve_service_checker_classes([], task_metadata=(1, 3))
+    task_two = run_scheduled_checks._resolve_service_checker_classes([], task_metadata=(2, 3))
+
+    assert [checker.service_key for checker in task_zero] == ["alpha", "bravo", "charlie"]
+    assert [checker.service_key for checker in task_one] == ["delta", "echo"]
+    assert [checker.service_key for checker in task_two] == ["foxtrot", "golf"]
 
 
 @pytest.mark.asyncio