My use-case may be somewhat unique: I have a central server, running Windows, that I’d like to keep reasonably-secure and accessible to coworkers using an efficient (meshed) VPN like zerotier, and one of our constraints is that we are using it over cellular hotspots (or tethering) at one end – and we move around some while we use it.
Today, myself and an associate were in the field in a medium-sized semi-rural factory doing some programming of [stuff] using zerotier on our laptops to connect to that central server via our cell phones’ wifi hotspots.
And that worked well, except when it didn’t. In areas of poor signal strength [or signal quality], cell phones can jump betwixt different IP addresses: Most people never notice it because things like streaming can often work just fine in this network environment, but we weren’t streaming: We just needed connectivity to do our work as we walked around doing various things to various bits of fixed, non-networked equipment.
This intermittent nature seems to often confuse zerotier for minutes at a time, and that sucks for tasks that also only take minutes: It can double or triple the amount of time it takes to complete a task as we wait for zerotier (when combined with the underlying cellular SNAFU) to stabilize as we move about.
So, I’m looking for workarounds that promote usability and/or stability in less-than-stellar networking environments.
Let’s start with stability. I assume that Zerotier does keepalives or similar, and that after a certain number of them are missed or different then things reconfigure themselves, and that this is why zerotier takes minutes to get its feet back on the ground once the network changes. If this is the case, then: Would it be the case that more-frequent keepalives could help? If so, then: Is that possible? (Ideally, I’d like zerotier to hop around in the background just about as quickly as the local computer’s WAN connection also hops around – even if that has other expenses like idle overhead.)
I assume this could also be improved with port forwarding (or a static IP and a non-NAT firewall) instead of relying on NAT hole-punching, but the server in question is one of several zerotier devices on the LAN and I don’t see a way to tell zerotier to use a particular port for a particular device so this path seems to be very indeterminate – and besides, port forwarding seems to be actively discouraged. (Edit: In many cases, UPnP is more of a problem than it is a solution, and this is one of those cases.)
I might be barking up the wrong tree here, but I think that embodies my present thoughts on stability.
Meanwhile, usability: The Windows zerotier UI is rather spartan. It takes too many clicks (whereby the user polls the systems) to get an idea of what the status of those systems, like the VPN backend service, might be. That seems dumb: The UI, when prompted to do so, should poll its own backend without the user doing anything further, and present the user with a regularly-updated status screen. This would really help troubleshoot issues.
Right now, it goes like this:
clickety-click F, it’s not working – again.
clickety-click Are we there yet?
clickety-click Are we there yet?
clickety-click Let’s try to disconnect.
clickety-click Have we disconnected?
clickety-click We’re waiting for the Windows service
clickety-click Still waiting
clickety-click Oh nice, the UI responds again.
clickety-click I think we’re disconnected.
clickety-click Yes, we are disconnected.
clickety-click Let’s reconnect!
clickety-click Are we there yet?
clickety-clicking rinsing, repeating
clickety-click meh
That’s waaaaay too much clickety-clicking. We have a very feature-rich GUI at our disposal under Windows, and it seems like it should do the clicking/polling for user and display results that can be seen at a glance instead of working so hard (so many clicks!) to get a peek behind the curtain while all lucrative work ceases until the issue is resolved.
After all: Isn’t a reduction of repetitive, manual steps one of the prime benefits of using a computer?
Again, perhaps this is not a common use-case, but it is my use-case.
Other than that: Sure, we got the job done today. Thanks! Awesome! I just wish that aspects of it were speedier, and easier to observe and/or control.
What can we do to improve any of these aspects? Any ideas or discussion? (Am I to write my own UI that automates some of these repetitive tasks to improve usability?)
(We are a small-potatoes paid subscriber, but I didn’t make a ticket because I don’t feel that this kind of spitballing rises to the level that would merit a ticket.)