fix(sys-stk-front-proxy): probe the queued vhosts only on the OpenResty host

sys-stk-full includes the front proxy with run_once and delegate_to, so every set_fact inside is broadcast to all hosts of the play. A worker context queued the clearnet URL (its group_names lack svc-net-tor) over the manager's onion queue, and the deferred probe then ran on every host: workers and the NFS server, which serve nothing and cannot resolve .onion, answered -1 and failed the swarm deploy with 'OpenResty does not serve ... even after a restart' after a needless OpenResty restart. Only the stack host now queues a vhost, and only the first host of svc-prx-openresty probes the queue; an empty queue skips before the group is read.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
This commit is contained in:
kevinveenbirkenbachandClaude Opus 5.5 committed 2026-10-29 08:26:22 +01:00
1 parent 04257b5b76
commit aa88429c25
2 files changed
+4 -1

No files matched your search

@@ -1,6 +1,8 @@
---
- name: "🔍 Probe the vhosts this play queued"
when: front_proxy_deferred_probes | length > 0
when:
- front_proxy_deferred_probes | length > 0
- inventory_hostname == groups['svc-prx-openresty'][0]
block:
- name: "🔍 Check if the queued vhosts are reachable"
uri_probe_batch:
+1
View File
@@ -55,6 +55,7 @@
url: "{{ lookup('tls', front_proxy_domain, 'url.base') }}"
status_codes: "{{ lookup('applications') | accepted_status_codes(front_proxy_application_id, lookup('tls', front_proxy_domain, 'domain')) }}"
changed_when: false
when: IS_STACK_HOST | bool
- name: 📂 Include dependencies
include_role: