CI: Disable stack cost retrieval to stop job hang

This is a temporary workaround to address critical job hangs caused
by lftools openstack stack cost command lacking timeout handling.

Problem (IT-28614):
  The lftools stack cost command hangs indefinitely when:
  - Pricing API (pricing.vexxhost.net) is slow/unresponsive
  - OpenStack SDK queries for nested stacks take too long
  - Network operations lack timeout parameters

  This causes:
  - Jobs stuck at 'Retrieving stack cost for: <stack-name>'
  - Downstream jobs waiting on checkpoints indefinitely
  - Jenkins queue buildup (66+ jobs waiting reported)
  - Requires manual intervention to cancel stuck jobs

Solution:
  Disable stack cost retrieval by commenting out the lftools call
  and hardcode stack cost to 0 in the stack-cost file.

Known Regression:
  Stack costs will be reported as 0 in archived cost.csv files,
  losing cost tracking data for OpenStack resources. Instance-level
  costs are still collected via job-cost.sh pricing API calls.

Root Cause:
  lftools/openstack/stack.py cost() function lacks:
  - timeout parameter for urllib.request.urlopen() (line 87)
  - timeout handling for OpenStack SDK resource enumeration
  - graceful degradation on network failures

Next Steps:
  1. Implement timeout handling in lftools
  2. Once lftools is fixed, revert this workaround
  3. Create tracking issue for the regression (stack costs = 0)

This stopgap prioritizes job reliability over cost data collection.

Issue-ID: IT-28614
Change-Id: I9d054347e485f4843c7287aec49fb7aff0962e96
Signed-off-by: Anil Belur <abelur@linuxfoundation.org>

diff --git a/releasenotes/notes/disable-stack-cost-retrieval-6ab5fd84c23b994e.yaml b/releasenotes/notes/disable-stack-cost-retrieval-6ab5fd84c23b994e.yaml
new file mode 100644 (file)
index 0000000..b08d8a2
--- /dev/null
+++ b/releasenotes/notes/disable-stack-cost-retrieval-6ab5fd84c23b994e.yaml
@@ -0,0 +1,32 @@
+---
+fixes:
+  - |
+    **STOPGAP FIX:** Temporarily disable stack cost retrieval in
+    openstack-stack-delete.sh to prevent jobs from hanging indefinitely.
+
+    **Problem:** The lftools openstack stack cost command hangs without
+    timeout handling when querying the pricing API or OpenStack resources,
+    causing:
+
+    - Jobs stuck at "Retrieving stack cost" phase
+    - Subsequent jobs waiting on checkpoints
+    - Jenkins queue buildup and capacity issues
+
+    **Solution:** Stack cost is now hardcoded to 0 until lftools implements
+    proper timeout handling for network operations.
+
+    **Known Regression:** Stack costs will be reported as 0 in cost.csv
+    files, losing cost tracking data for OpenStack resources. Instance-level
+    costs from the pricing API are still collected via job-cost.sh.
+
+    **Root Cause:** lftools lacks timeout parameters for urllib.request.urlopen()
+    and OpenStack SDK operations in the stack cost calculation.
+
+    **Tracking:**
+
+    - Fixes: IT-28614 (Jobs hanging on stack cost retrieval)
+    - Regression: Stack costs always reported as 0 (needs lftools fix)
+    - Upstream: lftools needs timeout implementation for openstack stack cost command
+
+    **Next Steps:** Once lftools implements timeouts, this workaround should
+    be reverted to restore cost collection functionality.

diff --git a/shell/openstack-stack-delete.sh b/shell/openstack-stack-delete.sh
index 158dcfc..a04fdb3 100644 (file)
--- a/shell/openstack-stack-delete.sh
+++ b/shell/openstack-stack-delete.sh
@@ -21,13 +21,17 @@ lf-activate-venv --python python3 "lftools[openstack]" \
     python-openstackclient \
     urllib3~=1.26.15
 
-echo "INFO: Retrieving stack cost for: $OS_STACK_NAME"
-if ! lftools openstack --os-cloud "$OS_CLOUD" stack cost "$OS_STACK_NAME" > stack-cost; then
-    echo "WARNING: Unable to get stack costs, continuing anyway"
-    echo "total: 0" > stack-cost
-else
-    echo "DEBUG: Successfully retrieved stack cost: $(cat stack-cost)"
-fi
+# Disabled stack cost retrieval due to lftools hanging without timeout
+# See: OpenDaylight Jenkins issue with stuck jobs
+# echo "INFO: Retrieving stack cost for: $OS_STACK_NAME"
+# if ! lftools openstack --os-cloud "$OS_CLOUD" stack cost "$OS_STACK_NAME" > stack-cost; then
+#     echo "WARNING: Unable to get stack costs, continuing anyway"
+#     echo "total: 0" > stack-cost
+# else
+#     echo "DEBUG: Successfully retrieved stack cost: $(cat stack-cost)"
+# fi
+echo "INFO: Stack cost retrieval disabled, setting cost to 0"
+echo "total: 0" > stack-cost
 
 # Delete the stack even if the stack-cost script fails
 lftools openstack --os-cloud "$OS_CLOUD" stack delete "$OS_STACK_NAME" \