packages/system/cozystack-scheduler/examples/demo/00-helpers.sh (3)

91-97: Unusual stderr redirection in show_pods.

Line 95 uses 2>&1 >&2 which redirects stderr to the original stdout, then stdout to stderr. The intent seems to be sending all output to stderr, but 2>&1 captures stderr before stdout is redirected. This effectively sends stdout to stderr and stderr to the original stdout (likely also stderr in this context, so it works by coincidence).

A clearer approach:

♻️ Proposed simplification

 show_pods() {
     local namespace="$1"
     echo "" >&2
-    kubectl get pods -n "$namespace" -o wide 2>&1 >&2
+    kubectl get pods -n "$namespace" -o wide >&2 2>&1
     echo "" >&2
 }

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/system/cozystack-scheduler/examples/demo/00-helpers.sh` around lines
91 - 97, The show_pods function uses the confusing redirection "2>&1 >&2" for
the kubectl get pods call; replace that with a single clear redirect so both
stdout and stderr go to stderr (e.g., change the kubectl get pods -n
"$namespace" -o wide 2>&1 >&2 invocation to a single redirection such as kubectl
get pods -n "$namespace" -o wide >&2) to ensure all output is sent to stderr;
update the kubectl line inside the show_pods function accordingly.

---

62-89: Consider checking pod readiness, not just existence.

wait_for_pods only waits for pods to appear (count > 0), but doesn't verify they're actually running or ready. For demo purposes this may be intentional (you want to show the scheduling quickly), but the function name and comment suggest it waits for pods to "stabilize."

If stabilization is desired, consider adding a readiness check:

♻️ Optional enhancement for readiness check

     while true; do
         local count
         if [[ -n "$label" ]]; then
-            count=$(kubectl get pods -n "$namespace" -l "$label" --no-headers 2>/dev/null | wc -l)
+            count=$(kubectl get pods -n "$namespace" -l "$label" --field-selector=status.phase=Running --no-headers 2>/dev/null | wc -l)
         else
-            count=$(kubectl get pods -n "$namespace" --no-headers 2>/dev/null | wc -l)
+            count=$(kubectl get pods -n "$namespace" --field-selector=status.phase=Running --no-headers 2>/dev/null | wc -l)
         fi

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/system/cozystack-scheduler/examples/demo/00-helpers.sh` around lines
62 - 89, wait_for_pods currently only checks pod existence (count > 0) and not
readiness; change it to wait for pods to be Ready/stable by using kubectl wait
--for=condition=ready (or polling kubectl get pods -n "$namespace" -l "$label"
-o jsonpath='{range .items[*]}{.metadata.name}:{.status.phase}:{range
.status.conditions[*]}{.type}={.status};{end}{"\n"}{end}') and ensure all
matching pods report Ready=true (or .status.phase==Running) before breaking;
keep the same parameters (namespace, label, timeout) and update log messages
(log_info/log_warning) and return codes so the function returns success only
when pods are ready or returns 1 on timeout.

---

99-121: Use condition type filter instead of array index to avoid fragile status checks.

The jsonpath .status.conditions[0].status assumes the first condition is always the "Ready" condition. Kubernetes resources typically have multiple conditions (e.g., Initialized, Ready, ContainersReady), and order is not guaranteed. Filter by condition type instead:

♻️ Proposed fix to query by condition type

     while true; do
         local phase
-        phase=$(kubectl get redis "$name" -n "$namespace" -o jsonpath='{.status.conditions[0].status}' 2>/dev/null || echo "")
+        phase=$(kubectl get redis "$name" -n "$namespace" -o jsonpath='{.status.conditions[?(@.type=="Ready")].status}' 2>/dev/null || echo "")
         if [[ "$phase" == "True" ]]; then
             log_success "Redis $name is ready"
             break
         fi

This pattern aligns with how conditions are queried throughout the codebase (dashboard, packagesource types, and other operators).

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/system/cozystack-scheduler/examples/demo/00-helpers.sh` around lines
99 - 121, The wait_for_redis_ready function uses the fragile jsonpath
.status.conditions[0].status; change the kubectl jsonpath to select the
condition by type (e.g. .status.conditions[?(@.type=="Ready")].status) so you
explicitly read the Ready condition, keep the same variable name phase and the
same truth check, and preserve the timeout/sleep logic and fallback to empty
string on kubectl errors.

packages/system/cozystack-scheduler/examples/demo/09-spread-demo.sh (1)

67-69: Consider: Awk column extraction assumes fixed kubectl output format.

The awk '{print $7}' assumes the NODE column is always the 7th field in kubectl get pods -o wide output. While this is standard, using -o custom-columns or -o jsonpath would be more robust.

For a demo script, this is acceptable, but for production tooling, consider:

kubectl get pods -n "$NAMESPACE" -o jsonpath='{range .items[*]}{.spec.nodeName}{"\n"}{end}' | sort | uniq -c | sort -rn

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.

In `@packages/system/cozystack-scheduler/examples/demo/09-spread-demo.sh` around
lines 67 - 69, Replace the brittle awk column extraction used in the "kubectl
get pods -n \"$NAMESPACE\" -o wide" pipeline (the awk '{print $7}' reference)
with a robust jsonpath-based extraction of the pod's node name (use kubectl -o
jsonpath to read .spec.nodeName for each item), then pipe the results to sort |
uniq -c | sort -rn so the log_step "Pod count per node:" output no longer
depends on column positions.