fix(tasks): Add autoretry for too many simultaneous queries error in some tasks

ee/clickhouse/test/test_error.py

@@ -35,6 +35,12 @@
             "Query exceeds memory limits. Try reducing its scope by changing the time range.",
             241,
         ),
+        (
+            ServerException("Too many simultaneous queries. Maximum: 100.", code=202),
+            "CHQueryErrorTooManySimultaneousQueries",
+            "Code: 202.\nToo many simultaneous queries. Try again later.",
+            202,
+        ),
     ],
 )
 def test_wrap_query_error(error, expected_type, expected_message, expected_code):

posthog/errors.py

@@ -57,6 +57,10 @@ def wrap_query_error(err: Exception) -> Exception:
     # :TRICKY: Return a custom class for every code by looking up the short name and creating a class dynamically.
     if hasattr(err, "code"):
         meta = look_up_error_code_meta(err)
+
+        if meta.name in CLICKHOUSE_SPECIFIC_ERROR_LOOKUP:
+            return CLICKHOUSE_SPECIFIC_ERROR_LOOKUP[meta.name]
+
         name = f"CHQueryError{meta.name.replace('_', ' ').title().replace(' ', '')}"
         processed_error_class = ExposedCHQueryError if meta.user_safe else InternalCHQueryError
         message = meta.user_safe if isinstance(meta.user_safe, str) else err.message
@@ -71,6 +75,19 @@ def look_up_error_code_meta(error: ServerException) -> ErrorCodeMeta:
     return CLICKHOUSE_ERROR_CODE_LOOKUP[code]
 
 
+# Specific error classes we need
+# These exist here and are not dynamically created because they are used in the codebase.
+class CHQueryErrorTooManySimultaneousQueries(InternalCHQueryError):
+    pass
+
+
+CLICKHOUSE_SPECIFIC_ERROR_LOOKUP = {
+    "TOO_MANY_SIMULTANEOUS_QUERIES": CHQueryErrorTooManySimultaneousQueries(
+        "Too many simultaneous queries. Try again later.", code=202
+    ),
+}
+
+
 #
 # From https://github.com/ClickHouse/ClickHouse/blob/23.12/src/Common/ErrorCodes.cpp#L16-L622
 #

posthog/tasks/exporter.py

@@ -6,6 +6,7 @@
 from django.db import transaction
 
 from posthog import settings
+from posthog.errors import CHQueryErrorTooManySimultaneousQueries
 from posthog.models import ExportedAsset
 from posthog.tasks.utils import CeleryQueue
 
@@ -43,6 +44,10 @@
     ignore_result=False,
     time_limit=settings.ASSET_GENERATION_MAX_TIMEOUT_SECONDS,
     queue=CeleryQueue.EXPORTS.value,
+    autoretry_for=(CHQueryErrorTooManySimultaneousQueries,),
+    retry_backoff=1,
+    retry_backoff_max=3,
+    max_retries=3,
 )
 @transaction.atomic
 def export_asset(exported_asset_id: int, limit: Optional[int] = None) -> None:

posthog/tasks/tasks.py

@@ -9,6 +9,7 @@
 from prometheus_client import Gauge
 
 from posthog.cloud_utils import is_cloud
+from posthog.errors import CHQueryErrorTooManySimultaneousQueries
 from posthog.hogql.constants import LimitContext
 from posthog.metrics import pushed_metrics_registry
 from posthog.ph_client import get_ph_client
@@ -32,7 +33,18 @@ def redis_heartbeat() -> None:
     get_client().set("POSTHOG_HEARTBEAT", int(time.time()))
 
 
-@shared_task(ignore_result=True, queue=CeleryQueue.ANALYTICS_QUERIES.value, acks_late=True)
+@shared_task(
+    ignore_result=True,
+    queue=CeleryQueue.ANALYTICS_QUERIES.value,
+    acks_late=True,
+    autoretry_for=(
+        # Important: Only retry for things that might be okay on the next try
+        CHQueryErrorTooManySimultaneousQueries,
+    ),
+    retry_backoff=1,
+    retry_backoff_max=2,
+    max_retries=3,
+)
 def process_query_task(
     team_id: int,
     user_id: int,
@@ -550,7 +562,14 @@ def schedule_cache_updates_task() -> None:
     schedule_cache_updates()
 
 
-@shared_task(ignore_result=True)
+@shared_task(
+    ignore_result=True,
+    autoretry_for=(CHQueryErrorTooManySimultaneousQueries,),
+    retry_backoff=10,
+    retry_backoff_max=30,
+    max_retries=3,
+    retry_jitter=True,
+)
 def update_cache_task(caching_state_id: UUID) -> None:
     from posthog.caching.insight_cache import update_cache