You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Sep 26, 2021. It is now read-only.
I'd like to have one machine with Consul master instance and Swarm manager (like in #2302) and set of nodes each of which will run Swarm agent + own Consul instance connected with master.
Currently it is not possible to make out of the box, since IP of machine with Swarm manager is unknown (at least when using Vitrualbox driver), so I can't provide Consul IP to it because it should be at the same place.
So one way is to allow creating machine first, then manually install Consul on it and then run Swarm installation with Machine. Alternative way for this particular problem is to allow Consul installation to be completely automated using Machine out of the box.
For second approach we need to create machine with Consul AND Swarm created at the same time, so that master machine will be created with one command and then for all nodes IP of master would be treated as both Consul master and Swarm manager.
So basically, --swarm-discovery-consul will also run Consul container and insert proper consul://<consul_addr> argument for Swarm.
And if --swarm-master not specified - Consul master assumed to be where Swarm master is located.
P.S. This is not Consul-specific, similar keys might be added for etcd and any other by analogy.
The text was updated successfully, but these errors were encountered:
I, as well, would like the ability to swarm-ify an existing machine (without having to remove it first).
Something even more useful, I think, for container "provisioning" during docker-machine create would be the ability to inject compositions to be initialized. One assumes that the swarm containers themselves would be the default composition during the setup of a swarm node.
Yeah, I am not 100% sure, but my thinking on this is that to cover this and other modify-after-creation cases we should start eventually start supporting a config file, which would allow you to converge state on these types of things by calling some sort of apply-type command.
I'd like to have one machine with Consul master instance and Swarm manager (like in #2302) and set of nodes each of which will run Swarm agent + own Consul instance connected with master.
Currently it is not possible to make out of the box, since IP of machine with Swarm manager is unknown (at least when using Vitrualbox driver), so I can't provide Consul IP to it because it should be at the same place.
So one way is to allow creating machine first, then manually install Consul on it and then run Swarm installation with Machine. Alternative way for this particular problem is to allow Consul installation to be completely automated using Machine out of the box.
For second approach we need to create machine with Consul AND Swarm created at the same time, so that master machine will be created with one command and then for all nodes IP of master would be treated as both Consul master and Swarm manager.
How I see this in CLI interface:
So basically,
--swarm-discovery-consul
will also run Consul container and insert properconsul://<consul_addr>
argument for Swarm.And if
--swarm-master
not specified - Consul master assumed to be where Swarm master is located.P.S. This is not Consul-specific, similar keys might be added for etcd and any other by analogy.
The text was updated successfully, but these errors were encountered: