fix: [nomad-step-1] S1.2 — add lib/init/nomad/deploy.sh (dependency-ordered nomad job run + wait) (#841)

lib/init/nomad/deploy.sh

@@ -30,7 +30,7 @@ set -euo pipefail
 
 # ── Configuration ────────────────────────────────────────────────────────────
 SCRIPT_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
-REPO_ROOT="${REPO_ROOT:-$(cd "${SCRIPT_ROOT}/../../.." && pwd)}"
+REPO_ROOT="${REPO_ROOT:-$(cd "${SCRIPT_ROOT}/../.." && pwd)}"
 JOB_READY_TIMEOUT_SECS="${JOB_READY_TIMEOUT_SECS:-120}"
 
 DRY_RUN=0
@@ -96,14 +96,28 @@ _wait_job_running() {
         log "job '${job_name}' is now running"
         return 0
         ;;
-      complete)
+      complete|dead|failed)
+        # Check allocations for partial success
+        local allocs_running
+        allocs_running=$(printf '%s' "$status_json" \
+          | jq '[.Evaluations[].Allocations[]? | select(.Status == "running")] | length' 2>/dev/null) || allocs_running=0
+        local allocs_total
+        allocs_total=$(printf '%s' "$status_json" \
+          | jq '[.Evaluations[].Allocations[]? | length] | add' 2>/dev/null) || allocs_total=0
+
+        if [ "$allocs_running" -gt 0 ]; then
+          log "job '${job_name}' has ${allocs_running}/${allocs_total} allocations running"
+          # If not all running but some are, keep waiting
+          if [ "$allocs_running" -lt "$allocs_total" ]; then
+            sleep 5
+            elapsed=$((elapsed + 5))
+            continue
+          fi
+        fi
+
         log "job '${job_name}' reached terminal state: ${status}"
         return 0
         ;;
-      dead|failed)
-        log "job '${job_name}' reached terminal state: ${status}"
-        return 1
-        ;;
       *)
         log "job '${job_name}' status: ${status} (waiting...)"
         ;;
@@ -156,14 +170,11 @@ for job_name in "${JOBS[@]}"; do
   fi
 
   # 2. Check if already running (idempotency)
-  job_status_json=$(nomad job status -json "$job_name" 2>/dev/null || true)
-  if [ -n "$job_status_json" ]; then
-    current_status=$(printf '%s' "$job_status_json" | jq -r '.Status' 2>/dev/null || true)
-    if [ "$current_status" = "running" ]; then
+  job_status=$(nomad job status "$job_name" 2>/dev/null | head -1 || true)
+  if printf '%s' "$job_status" | grep -qi "running"; then
     log "${job_name} already running"
     continue
   fi
-  fi
 
   # 3. Run the job (idempotent registration)
   log "running: ${jobspec_path}"