If I have machine A, host XYZ, and machine B, and both machine A and B are using host XYZ as their default remote, can machine A run an image from host XYZ locally so that the container has access to its hardware and have machine B who is also using host XYZ as its default be able to see machine A’s instance that is running locally on machine A when machine B executes lxc list? Could --force-local help with this?
No, from what you’ve described, both A and B see whatever runs on XYZ.
You can run something like lxc launch XYZ:your-image local:foo, running that on A would download the image your-image from XYZ and create a local instance called foo, but B doesn’t have A as a remote so it wouldn’t see that instance.
It sounds like you either need to add all those systems to each other as remotes and then make sure you select the correct remote every time, or you may want to cluster the 3 machines together which would let you see all instances at once from any of the 3 systems, but that also comes with the requirement that all 3 have similar storage and network setup and that at least 2 remain online at any one time.
There are around 300 machines located around NA and a few in EU, just from sshing into them individually and pinging other members, there seems to be around 300-1000ms latency between them something else to note is that they have 5gb bandwidth limit per day, I look at Do all LXD Cluster member nodes have to be on the same LAN - #2 by tomp
and
Lxc list is slow when getting ipv4 addresses - #4 by tomp
and I feel like my setup may not be a good candidate for clustering, I wouldn’t really mind lxc list taking 3-5 minutes to finish though but I’m not sure what else to expect, can you tell me how plausible clustering would be for this kind of infrastructure or if I should just stick to having the one remote that I pull images from onto the machines and run them on their local environment?