Prevent "vault" service to be restarted on update

[legal90]

Vault server restart is even more harmful than consul restart (related: sous-chefs/consul#288).

Vault server restart means that Vault becomes sealed and stops process incoming requests. So, regardless of the reason, it should be done only by an operator.

[davidsayers]

Would it not make more sense to have this a configurable option instead of forcing users one way or the other?

[Ginja]

Good idea, @legal90. But I also agree with @davidsayers. Would you mind adding an attribute for toggling this? It would give power users the flexibility to come up with their own restart methodology.

default['hashicorp-vault']['service']['restart_on_update'] = false

Also, the Vault service will still restart if one of the config options are changed. Perhaps wrap this property in a conditional with the above attribute?

https://github.com/johnbellone/vault-cookbook/blob/master/recipes/default.rb#L18

[Ginja]

Actually, I think this can be overridden with the correct poise-service attribute defined:

override['poise-service']['vault']['restart_on_update'] = false

@coderanger, is that right?

[legal90]

Would you mind adding an attribute for toggling this?

OK, I'll do this.
UPDATE: PR was merged as is.

Also, the Vault service will still restart if one of the config options are changed. Perhaps wrap this property in a conditional with the above attribute?

There is reload, not restart. The first one is just sending a SIGHUP, which is safe action and doesn't cause Vault to be sealed.

[Ginja]

Ah, good catch! Not enough ☕.

[johnbellone]

This makes sense to me.

[coderanger]

I would probably use the never_restart restart option rather than this.

service.options never_restart: true

All providers should support that option to make the restart action a no-op.

[coderanger]

And yes, restart_on_update can be overridden via an option too but that only covers sending a restart notification when the service definition itself changes.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.