fix: retain healthy terminal runners

package/gitea-runner-controller/webhook.sh

@@ -155,6 +155,7 @@ gcr_alloc() {
         vm_id="$(gcr_record_field "$reused" vm_id)"
         vm_name="$(gcr_record_field "$reused" vm_name)"
         if gcr_vm_runner_service "$vm_id" start \
+            && gcr_vm_runner_service "$vm_id" health \
             && gcr_gitea_runner_disabled "$repo" "$vm_name" false; then
             reused="$(gcr_record_get "$job_id" "$attempt")"
             reused="$(printf '%s' "$reused" | jq -c '.bootstrapped = true | del(.reused_vm)')"
@@ -247,54 +248,14 @@ gcr_deallocate() {
             ;;
     esac
 
-    vm_id="$(gcr_record_field "$rec" vm_id)"
+    finish_status=0
-    if [ "$new_status" = "completed:success" ] \
+    gcr_vm_finish_terminal "$job_id" "$attempt" "$rec" "$new_status" webhook \
-        && idle_rec="$(gcr_record_idle_json "$rec")"; then
+        || finish_status="$?"
-        if ! gcr_lock_acquire idle-pool; then
     gcr_lock_release "$key"
-            return 0
+    case "$finish_status" in
-        fi
+        0|2) return 0 ;;
-        runner_name="$(gcr_record_field "$rec" vm_name)"
+        *) return "$finish_status" ;;
-        if ! gcr_gitea_runner_disabled "$(gcr_record_field "$rec" repo)" "$runner_name" true \
-            || ! gcr_vm_runner_service "$vm_id" stop; then
-            gcr_lock_release idle-pool
-            if gcr_vm_cleanup_start "$job_id" "$attempt" "$rec" idle-stop-failed false; then
-                gcr_event "vm-destroyed" "$job_id" \
-                    "{\"vm_id\":$vm_id,\"reason\":\"idle-stop-failed\"}"
-            else
-                gcr_event "vm-cleanup-pending" "$job_id" \
-                    "{\"vm_id\":$vm_id,\"reason\":\"idle-stop-failed\"}"
-            fi
-            gcr_lock_release "$key"
-            return 0
-        fi
-        gcr_record_put "$job_id" "$attempt" "$idle_rec"
-        idle_expires="$(gcr_record_field "$idle_rec" idle_expires_at)"
-        gcr_lock_release idle-pool
-        gcr_lock_release "$key"
-        gcr_event "vm-idle" "$job_id" "{\"vm_id\":$vm_id,\"expires_at\":$idle_expires}"
-        return 0
-    fi
-
-    if [ -n "$vm_id" ] && [ "$vm_id" != "null" ] && [ "$vm_id" != "0" ]; then
-        case "$new_status" in
-            completed:success|completed:cancelled|completed:skipped) ;;
-            completed:*)
-                ip="$(gcr_vm_public_ip "$vm_id" || true)"
-                gcr_vm_collect_diagnostics "$vm_id" "$ip" "$job_id" "$new_status" || true
-                ;;
     esac
-        if gcr_vm_cleanup_start "$job_id" "$attempt" "$rec" "$new_status" false; then
-            gcr_event "vm-destroyed" "$job_id" \
-                "{\"vm_id\":$vm_id,\"reason\":\"$new_status\"}"
-        else
-            gcr_event "vm-cleanup-pending" "$job_id" \
-                "{\"vm_id\":$vm_id,\"reason\":\"$new_status\"}"
-        fi
-    else
-        gcr_record_del "$job_id" "$attempt"
-    fi
-    gcr_lock_release "$key"
 }
 
 gcr_mark_in_progress() {

test/package/gitea-runner-controller/test/webhook-diagnostics.sh

@@ -28,7 +28,9 @@ gcr_vm_runner_service() {
   printf 'runner %s vm=%s\n' "$2" "$1" >> "$calls"
 }
 
-gcr_gitea_runner_disabled() { :; }
+gcr_gitea_runner_disabled() {
+  printf 'disabled runner=%s value=%s\n' "$2" "$3" >> "$calls"
+}
 
 record_success='{"job_id":"201","run_attempt":"1","repo":"hinterland/hearth","label":"gross-nix-x86-perf","created_at":"0","ttl_min":480,"vm_id":51,"vm_name":"gcr-201-1","bootstrapped":true,"status":"vm_active"}'
 record_failure='{"job_id":"202","run_attempt":"1","repo":"hinterland/hearth","label":"gross-nix-x86-perf","created_at":"1","ttl_min":480,"vm_id":52,"vm_name":"gcr-202-1","bootstrapped":true,"status":"vm_active"}'
@@ -38,8 +40,14 @@ gcr_record_put 202 1 "$record_failure"
 gcr_deallocate 201 1 completed:success
 gcr_deallocate 202 1 completed:failure
 
-grep -q 'destroy vm=52' "$calls"
 grep -q 'diag vm=52 ip=192.0.2.52 job=202 reason=completed:failure' "$calls"
+grep -q 'runner health vm=52' "$calls"
+grep -q 'disabled runner=gcr-202-1 value=true' "$calls"
+grep -q 'runner stop vm=52' "$calls"
+test "$(grep -E '^(diag vm=52|runner health vm=52|disabled runner=gcr-202-1|runner stop vm=52)' "$calls")" = 'diag vm=52 ip=192.0.2.52 job=202 reason=completed:failure
+runner health vm=52
+disabled runner=gcr-202-1 value=true
+runner stop vm=52'
 if grep -q 'diag vm=51' "$calls"; then
   printf 'success webhook should not collect diagnostics\n' >&2
   exit 1
@@ -49,12 +57,36 @@ if grep -q 'destroy vm=51' "$calls"; then
   printf 'successful webhook VM should remain idle until billing boundary\n' >&2
   exit 1
 fi
+if grep -q 'destroy vm=52' "$calls"; then
+  printf 'failed bootstrapped webhook VM should remain idle until billing boundary\n' >&2
+  exit 1
+fi
 jq -e 'select(.status == "idle_vm" and .idle_expires_at == 3600)' \
   "$(gcr_record_path 201 1)" >/dev/null
+jq -e 'select(.status == "idle_vm" and .idle_expires_at == 3601)' \
+  "$(gcr_record_path 202 1)" >/dev/null
+gcr_idle_record_usable "$(gcr_record_get 202 1)"
 idle_once="$(gcr_record_get 201 1)"
+gcr_record_del 201 1
+gcr_lock_acquire "$(gcr_alloc_key 205 1)"
+gcr_claim_idle 205 1 hinterland/hearth gross-nix-x86-perf
+gcr_lock_release "$(gcr_alloc_key 205 1)"
+jq -e 'select(.job_id == "205" and .vm_id == 52 and .status == "pending_vm" and .reused_vm == true)' \
+  "$(gcr_record_path 205 1)" >/dev/null
+gcr_record_del 205 1
+gcr_record_put 201 1 "$idle_once"
 gcr_deallocate 201 1 completed:success
 test "$(gcr_record_get 201 1)" = "$idle_once"
-test ! -e "$(gcr_record_path 202 1)"
+
+# Cancelled terminal jobs use same retention policy without failure diagnostics.
+record_cancelled='{"job_id":"207","run_attempt":"1","repo":"hinterland/hearth","label":"nix","created_at":"1","ttl_min":480,"vm_id":57,"vm_name":"gcr-207-1","bootstrapped":true,"status":"vm_active"}'
+gcr_record_put 207 1 "$record_cancelled"
+gcr_deallocate 207 1 completed:cancelled
+test "$(gcr_record_field "$(gcr_record_get 207 1)" status)" = idle_vm
+if grep -q 'diag vm=57' "$calls" || grep -q 'destroy vm=57' "$calls"; then
+  printf 'healthy cancelled VM must be retained without failure diagnostics\n' >&2
+  exit 1
+fi
 
 calls_ip_fail="$GCR_STATE_DIR/calls-ip-fail"
 calls="$calls_ip_fail"
@@ -64,20 +96,33 @@ gcr_record_put 203 1 "$record_ip_fail"
 gcr_vm_public_ip() {
   return 1
 }
+gcr_vm_runner_service() {
+  printf 'runner %s vm=%s\n' "$2" "$1" >> "$calls"
+  return 1
+}
 
 gcr_deallocate 203 1 completed:failure
 
 grep -q 'diag vm=53 ip= job=203 reason=completed:failure' "$calls_ip_fail"
+grep -q 'runner health vm=53' "$calls_ip_fail"
 grep -q 'destroy vm=53' "$calls_ip_fail"
 test ! -e "$(gcr_record_path 203 1)"
+if grep -q 'disabled runner=gcr-203-1' "$calls_ip_fail"; then
+  printf 'unhealthy terminal runner must not enter idle shutdown path\n' >&2
+  exit 1
+fi
 
-# Terminal webhook retains ownership and capacity until DELETE succeeds.
+# Runner teardown failure destroys and retains ownership until DELETE succeeds.
 record_delete_fail='{"job_id":"204","run_attempt":"1","repo":"hinterland/hearth","label":"nix","created_at":"1","ttl_min":480,"vm_id":54,"vm_name":"gcr-204-1","bootstrapped":true,"status":"vm_active"}'
 gcr_record_put 204 1 "$record_delete_fail"
 gcr_vm_destroy() {
   printf 'destroy-failed vm=%s\n' "$1" >> "$calls_ip_fail"
   return 1
 }
+gcr_vm_runner_service() {
+  printf 'runner %s vm=%s\n' "$2" "$1" >> "$calls"
+  [ "$2" != stop ]
+}
 gcr_deallocate 204 1 completed:failure
 cleanup_rec="$(gcr_record_get 204 1)"
 test "$(gcr_record_field "$cleanup_rec" status)" = cleanup_pending
@@ -94,6 +139,19 @@ test "$(gcr_count_active)" = 0
 grep -q '^destroy-failed vm=54$' "$calls_ip_fail"
 grep -q '^destroy-retry vm=54$' "$calls_ip_fail"
 
+# Unbootstrapped terminal VMs are never retained.
+gcr_vm_runner_service() { printf 'unexpected-runner %s\n' "$1" >> "$calls_ip_fail"; }
+gcr_vm_destroy() { printf 'destroy-unbootstrapped vm=%s\n' "$1" >> "$calls_ip_fail"; }
+record_unbootstrapped='{"job_id":"206","run_attempt":"1","repo":"hinterland/hearth","label":"nix","created_at":"1","ttl_min":480,"vm_id":56,"vm_name":"gcr-206-1","bootstrapped":false,"status":"pending_vm"}'
+gcr_record_put 206 1 "$record_unbootstrapped"
+gcr_deallocate 206 1 completed:failure
+grep -q '^destroy-unbootstrapped vm=56$' "$calls_ip_fail"
+test ! -e "$(gcr_record_path 206 1)"
+if grep -q '^unexpected-runner 56$' "$calls_ip_fail"; then
+  printf 'unbootstrapped VM must bypass idle teardown\n' >&2
+  exit 1
+fi
+
 # Gitea emits zero-based run_attempt values for initial workflow jobs.
 gcr_read_request() {
   gcr_hdr_event_type=workflow_job