From 2bada2d89c08b5be212bcad39ce04bacfc4f9357 Mon Sep 17 00:00:00 2001
From: Michael Vasseur <14887731+vmcj@users.noreply.github.com>
Date: Tue, 17 Sep 2024 13:25:53 +0200
Subject: [PATCH] Make sure tune_cpu gets started

In the past we could (and did) fail somewhere in the playbook which
would result in not handling this notified handler. This also meant that
we would not copy the file again on the next run, so didn't kick the
handler and also didn't hit the handler on the next run.

The disadvantage here is we might hit other handlers more often if next
roles want to hit the same handler.
---
 provision-contest/ansible/roles/judgedaemon/tasks/main.yml | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/provision-contest/ansible/roles/judgedaemon/tasks/main.yml b/provision-contest/ansible/roles/judgedaemon/tasks/main.yml
index 74001b70..e585b7e2 100644
--- a/provision-contest/ansible/roles/judgedaemon/tasks/main.yml
+++ b/provision-contest/ansible/roles/judgedaemon/tasks/main.yml
@@ -102,6 +102,10 @@
   notify:
     - Enable and restart tune_cpu
 
+- name: Enable and restart tune_cpu (flushing as this should always happen even if we fail further)
+  # We need a daemon-reload, so we loose idempotency if we start the service in the next block
+  ansible.builtin.meta: flush_handlers
+
 - name: Copy judgedaemon related systemd unit files
   copy:
     remote_src: true