Handle previously failed indexing jobs (#206)

Co-authored-by: Josh Bradley <[email protected]>

Author: clarked-msft

backend/indexing-job-manager-template.yaml

@@ -12,7 +12,6 @@ spec:
   schedule: "*/5 * * * *"
   jobTemplate:
     spec:
-      ttlSecondsAfterFinished: 30
       template:
         metadata:
           labels:

backend/indexing-job-template.yaml

@@ -26,8 +26,8 @@ spec:
         imagePullPolicy: Always
         resources:
           requests:
-            cpu: "6"
-            memory: "56Gi"
+            cpu: "5"
+            memory: "36Gi"
           limits:
             cpu: "8"
             memory: "64Gi"

backend/manage-indexing-jobs.py

@@ -88,7 +88,8 @@ def list_k8s_jobs(namespace: str) -> list[str]:
     jobs = batch_v1.list_namespaced_job(namespace=namespace)
     job_list = []
     for job in jobs.items:
-        job_list.append(job.metadata.name)
+        if job.metadata.name.startswith("indexing-job-") and job.status.active:
+            job_list.append(job.metadata.name)
     return job_list
 
 
@@ -124,7 +125,7 @@ def main():
                 )
                 pipelinejob = PipelineJob()
                 pipeline_job = pipelinejob.load_item(item["sanitized_index_name"])
-                pipeline_job["status"] = PipelineJobState.FAILED.value
+                pipeline_job.status = PipelineJobState.FAILED
             else:
                 print(
                     f"Indexing job for '{item['human_readable_index_name']}' already running. Will not schedule another. Exiting..."