Fix initial checkpoint save

backend/onyx/background/celery/tasks/indexing/tasks.py

@@ -367,21 +367,23 @@ def monitor_ccpair_indexing_taskset(
 
     redis_connector_index.reset()
 
-    # mark the CC Pair as `ACTIVE` if it's not already
+    # mark the CC Pair as `ACTIVE` if the attempt was a success and the
+    # CC Pair is not active not already
+    # This should never technically be in this state, but we'll handle it anyway
+    index_attempt = get_index_attempt(db_session, payload.index_attempt_id)
+    index_attempt_is_successful = index_attempt and index_attempt.status.is_successful()
     if (
-        # it should never technically be in this state, but we'll handle it anyway
-        cc_pair.status == ConnectorCredentialPairStatus.SCHEDULED
+        index_attempt_is_successful
+        and cc_pair.status == ConnectorCredentialPairStatus.SCHEDULED
         or cc_pair.status == ConnectorCredentialPairStatus.INITIAL_INDEXING
     ):
         cc_pair.status = ConnectorCredentialPairStatus.ACTIVE
         db_session.commit()
 
     # if the index attempt is successful, clear the repeated error state
-    if cc_pair.in_repeated_error_state:
-        index_attempt = get_index_attempt(db_session, payload.index_attempt_id)
-        if index_attempt and index_attempt.status.is_successful():
-            cc_pair.in_repeated_error_state = False
-            db_session.commit()
+    if cc_pair.in_repeated_error_state and index_attempt_is_successful:
+        cc_pair.in_repeated_error_state = False
+        db_session.commit()
 
 
 @shared_task(

backend/onyx/background/indexing/checkpointing_utils.py

@@ -160,11 +160,6 @@ def get_latest_valid_checkpoint(
         f"{latest_valid_checkpoint_candidate.id}. Previous checkpoint: "
         f"{previous_checkpoint}"
     )
-    save_checkpoint(
-        db_session=db_session,
-        index_attempt_id=latest_valid_checkpoint_candidate.id,
-        checkpoint=previous_checkpoint,
-    )
     return previous_checkpoint
 
 

backend/onyx/background/indexing/run_indexing.py

@@ -423,6 +423,14 @@ def _run_indexing(
                     connector=connector_runner.connector,
                 )
 
+            # save the initial checkpoint to have a proper record of the
+            # "last used checkpoint"
+            save_checkpoint(
+                db_session=db_session_temp,
+                index_attempt_id=index_attempt_id,
+                checkpoint=checkpoint,
+            )
+
             unresolved_errors = get_index_attempt_errors_for_cc_pair(
                 cc_pair_id=ctx.cc_pair_id,
                 unresolved_only=True,