I want to restart a service (haproxy) on a machine based on changes executed on another state. This would tipically be easy with onchanges but I have a particular case. Here's the state:
include:
- linux.v1_0
(...)
ftp.service:
service.running:
- reload: True
- watch:
- <what to use here?>
I've seen people saying to use the watch in order to restart. Thing is, the linux.v1_0 state included will make some changes and I want to restart the haproxy service only anf if changes are actually applied by the state. I would like to avoid to explicitly use systemctl restart haproxy.
The linux.v1_0 state has the following:
{% set ssl = pillar.get("c") %}
{% for domain, cert in ssl.items() %}
cert.ssl.crt.{{ domain }}:
file.managed:
- name: etc/ssl/{{ domain }}
(... more stuff ...)
{% endfor %}
So the task ID itself can vary from various domains, for example cert.ssl.crt.pt or cert.ssl.crt.es.
On the original state, I want the ftp service to restart ONLY when there are changes on the cert.ssl.crt.{{ domain }}task on the linux.v1_0 state.
To clarify, Salt requisites work on state ID. It does not matter whether a particular state ID is in found in included SLS or in-line. This can be checked with state.show_sls.
With respect to the dynamic state IDs getting generated, we can use the _in variant of the watch requisite, i.e. watch_in.
Like so:
{% for domain, cert in ssl.items() %}
cert.ssl.crt.{{ domain }}:
file.managed:
- name: etc/ssl/{{ domain }}
(... more stuff ...)
- watch_in:
service: ftp.service
{% endfor %}
While it will work for this requirement, any other SLS that includes linux.v1_0 will have to define ftp.service:. Please try to keep linux.v1_0 self-sufficient with related service restarts/reloads if possible.