Feature branch sync - pub/q2_upgrade to staging

build_stream/orchestrator/common/result_poller.py

@@ -362,6 +362,11 @@ def _on_result_received(self, result: PlaybookResult) -> None:
                 # S12: On restart failure, still persist node_results.json
                 if result.stage_name == "restart":
                     self._on_restart_completed(result)
+                    self._on_restart_failure(result)
+
+                # On deploy failure, mark ImageGroup FAILED
+                if result.stage_name == "deploy":
+                    self._on_deploy_failure(result)
 
                 # On validate failure, mark ImageGroup FAILED
                 if result.stage_name == "validate":
@@ -968,3 +973,49 @@ def _on_deploy_failure(self, result: PlaybookResult) -> None:
                 job_id=str(result.job_id),
                 exc_info=True,
             )
+
+    def _on_restart_failure(self, result: PlaybookResult) -> None:
+        """Transition ImageGroup from RESTARTING to FAILED on restart failure."""
+        if self._image_group_repo is None:
+            log_secure_info(
+                "warning",
+                f"ImageGroup repo not available; skipping restart failure "
+                f"update for job={result.job_id}",
+                job_id=str(result.job_id),
+            )
+            return
+
+        try:
+            image_group = self._image_group_repo.find_by_job_id(
+                JobId(str(result.job_id))
+            )
+            if image_group is None:
+                log_secure_info(
+                    "error",
+                    f"Restart failure callback: No ImageGroup found for job={result.job_id}.",
+                    job_id=str(result.job_id),
+                )
+                return
+
+            self._image_group_repo.update_status(
+                image_group_id=image_group.id,
+                new_status=ImageGroupStatus.FAILED,
+            )
+
+            if hasattr(self._image_group_repo, 'session'):
+                self._image_group_repo.session.commit()
+
+            log_secure_info(
+                "warning",
+                f"Restart FAILED for job={result.job_id}. "
+                f"ImageGroup '{image_group.id}' -> FAILED.",
+                job_id=str(result.job_id),
+            )
+        except Exception as exc:  # pylint: disable=broad-except
+            log_secure_info(
+                "error",
+                "Failed to update ImageGroup status on restart "
+                f"failure for job={result.job_id}: {exc}",
+                job_id=str(result.job_id),
+                exc_info=True,
+            )

common/library/module_utils/local_repo/software_utils.py

@@ -38,6 +38,7 @@
     CSV_COLUMNS,
     SOFTWARE_CONFIG_SUBDIR,
     DEFAULT_STATUS_FILENAME,
+    STATUS_CSV_HEADER,
     RPM_LABEL_TEMPLATE,
     RHEL_OS_URL,
     SOFTWARES_KEY,
@@ -853,6 +854,16 @@ def check_csv_existence(path):
 
 def read_status_csv(csv_path):
     """Reads the status.csv file and returns a list of row dictionaries."""
+    # Ensure file has valid header before reading
+    if os.path.exists(csv_path) and os.path.getsize(csv_path) > 0:
+        with open(csv_path, 'r', encoding='utf-8') as file:
+            lines = file.readlines()
+            if lines and lines[0].strip() != STATUS_CSV_HEADER.strip():
+                # Header missing or invalid - prepend header to existing data
+                with open(csv_path, 'w', encoding='utf-8') as wfile:
+                    wfile.write(STATUS_CSV_HEADER)
+                    wfile.writelines(lines)
+
     with open(csv_path, mode='r', newline='', encoding='utf-8') as file:
         reader = csv.DictReader(file)
         return [row for row in reader]

common/library/modules/parallel_tasks.py

@@ -160,9 +160,19 @@ def determine_function(
 
         # Construct the status file path using DEFAULT_STATUS_FILENAME.
         status_file = os.path.join(csv_file_path, DEFAULT_STATUS_FILENAME)
+
+        # Ensure file exists with valid header
         if not os.path.exists(status_file) or os.stat(status_file).st_size == 0:
             with open(status_file, 'w', encoding="utf-8") as file:
                 file.write(STATUS_CSV_HEADER)
+        else:
+            with open(status_file, 'r', encoding="utf-8") as file:
+                lines = file.readlines()
+                if lines and lines[0].strip() != STATUS_CSV_HEADER.strip():
+                    # Header missing or invalid - prepend header to existing data
+                    with open(status_file, 'w', encoding="utf-8") as wfile:
+                        wfile.write(STATUS_CSV_HEADER)
+                        wfile.writelines(lines)
 
 
         task_type = task.get("type")

examples/powerscale_reference_files/values.yaml → examples/powerscale_reference_files/CSI_driver/values.yaml

@@ -2,35 +2,35 @@
 ########################
 # version: version of this values file
 # Note: Do not change this value
-version: "v2.16.0"
+version: "v2.17.0"
 
 images:
   # "driver" defines the container image, used for the driver container.
   driver:
-    image: quay.io/dell/container-storage-modules/csi-isilon:v2.16.0
+    image: quay.io/dell/container-storage-modules/csi-isilon:v2.17.0
   # CSI sidecars
   attacher:
-    image: registry.k8s.io/sig-storage/csi-attacher:v4.10.0
+    image: registry.k8s.io/sig-storage/csi-attacher:v4.11.0
   provisioner:
-    image: registry.k8s.io/sig-storage/csi-provisioner:v6.1.0
+    image: registry.k8s.io/sig-storage/csi-provisioner:v6.2.0
   snapshotter:
-    image: registry.k8s.io/sig-storage/csi-snapshotter:v8.4.0
+    image: registry.k8s.io/sig-storage/csi-snapshotter:v8.5.0
   resizer:
-    image: registry.k8s.io/sig-storage/csi-resizer:v2.0.0
+    image: registry.k8s.io/sig-storage/csi-resizer:v2.1.0
   registrar:
-    image: registry.k8s.io/sig-storage/csi-node-driver-registrar:v2.15.0
+    image: registry.k8s.io/sig-storage/csi-node-driver-registrar:v2.16.0
   healthmonitor:
-    image: registry.k8s.io/sig-storage/csi-external-health-monitor-controller:v0.16.0
+    image: registry.k8s.io/sig-storage/csi-external-health-monitor-controller:v0.17.0
 
   # CSM sidecars
   replication:
-    image: quay.io/dell/container-storage-modules/dell-csi-replicator:v1.14.0
+    image: quay.io/dell/container-storage-modules/dell-csi-replicator:v1.15.0
   podmon:
-    image: quay.io/dell/container-storage-modules/podmon:v1.15.0
+    image: quay.io/dell/container-storage-modules/podmon:v1.16.0
   authorization:
-    image: quay.io/dell/container-storage-modules/csm-authorization-sidecar:v2.4.0
+    image: quay.io/dell/container-storage-modules/csm-authorization-sidecar:v2.5.0
   metadataretriever:
-    image: quay.io/dell/container-storage-modules/csi-metadata-retriever:v1.13.0
+    image: quay.io/dell/container-storage-modules/csi-metadata-retriever:v1.14.0
 
 # CSI driver log level
 # Allowed values: "error", "warn"/"warning", "info", "debug"
@@ -119,7 +119,7 @@ controller:
   # the Kubernetes release.
   # Allowed values: n, where n > 0
   # Default value: None
-  controllerCount: 2
+  controllerCount: 1
 
   # volumeNamePrefix: Prefix of PersistentVolume names created
   # Allowed values: string
@@ -184,15 +184,15 @@ controller:
     #   true: enable volume expansion feature(install resizer sidecar)
     #   false: disable volume snapshot feature(do not install resizer sidecar)
     # Default value: None
-    enabled: true
+    enabled: false
 
   healthMonitor:
     # enabled: Enable/Disable health monitor of CSI volumes- volume status, volume condition
     # Allowed values:
     #   true: enable checking of health condition of CSI volumes
     #   false: disable checking of health condition of CSI volumes
     # Default value: None
-    enabled: false
+    enabled: true
 
     # interval: Interval of monitoring volume health condition
     # Allowed values: Number followed by unit of time (s,m,h)
@@ -301,7 +301,7 @@ node:
     #   true: enable checking of health condition of CSI volumes
     #   false: disable checking of health condition of CSI volumes
     # Default value: None
-    enabled: false
+    enabled: true
 
 ## PLATFORM ATTRIBUTES
 ######################