Make waiting time for --tags=start configurable

Fixes https://github.com/spantaleev/matrix-docker-ansible-deploy/issues/1062

roles/matrix-common-after/defaults/main.yml

@@ -0,0 +1,16 @@
+# Specifies how long to wait between starting systemd services and checking if they're started.
+#
+# A too low value may lead to a failure, as services may not have enough time to start and potentially fail.
+#
+# A value higher than 30 seconds (or any multiple of that) may also not work well, because a failing systemd service
+# auto-restarts after 30 seconds (`RestartSec=30` in systemd service files).
+# Checking if a service is running right after it had potentially restarted in such a way will lead us to
+# thinking it's running, while it's merely starting again (and likely to fail again, given that it already did once).
+#
+# All of the services we manage are also started sequentially, which in itself can take a long time.
+# There may be a ~10 second (or even larger) interval between starting the first service and starting the last one.
+# This makes it even harder to pick a correct value. Such a 10 second gap and a waiting time of 20 seconds will
+# put us right at the "dangerous" 30-second mark.
+#
+# We can try to measure this gap and adjust our waiting time accordingly, but we currently don't.
+matrix_common_after_systemd_service_start_wait_for_timeout_seconds: 15

roles/matrix-common-after/tasks/start.yml

@@ -30,7 +30,7 @@
 # as we may run into systemd's automatic restart logic retrying the service.
 - name: Wait a bit, so that services can start (or fail)
   wait_for:
-    timeout: 15
+    timeout: "{{ matrix_common_after_systemd_service_start_wait_for_timeout_seconds }}"
   delegate_to: 127.0.0.1
   become: false