Keep internal apps up.
Even when a machine goes down.
Run an app on two machines under one name. Everyone uses the same one; if it stops, they all move to the other on their own. No load balancer, no DNS changes, nothing to update on anyone's device.
A machine that stops checking in is counted offline after 90 seconds. Then every device moves to the backup on its own. Same name, same address: nothing to change anywhere.
Three steps.
Nothing else to run.
Give the app a name, run it on more than one machine, and approve them. No load balancer or extra server to look after.
Give your app a name
In the dashboard, add a service and the port it uses. It gets its own name and its own address, separate from any one machine.
address 100.100.0.4
port tcp:443
Run it on two machines
On each machine that runs the app, one command says: this machine serves wiki, and here's where the app is listening.
--tcp=443 127.0.0.1:8080
Serving svc:wiki on port 443 from app-1.
Approve them, and you're done
A machine only serves once an admin approves it under Services in the dashboard. From then on, people just open wiki.
✓ app-2 · approved · backup
One name.
Machines behind it.
The name and address belong to the app, not to a machine. Add, swap or retire machines behind it, and nobody has to update a bookmark.
- Everyone goes to the same machine
- A backup is always ready
- Retire old machines without telling anyone
- 1app-1online · serving since MarchServing
- 2app-2online · serving since MayBackup
- 3app-3offlineSkipped
Machines that are online come first, then the one that has served the longest. Every device uses the same order, so everyone lands on the same machine.
- Name and address, whichever machine runs it
- 1
- Without a check-in before a machine counts as offline
- 90 s
- How often the network checks who's online
- 10 s
- Settings to change on anyone's device
- 0
The apps your team can't be without.
Kept running.
Anything people open every day, and notice the moment it's down.
Internal wiki or docs
The page your team checks ten times a day.
Dashboards and admin tools
Monitoring, back offices, anything behind a login.
Git and build servers
Keep pushes and builds working while you patch one.
Anything on a TCP port
Databases, APIs, SSH. If it listens on a port, it works.
Questions about uptime.
Something we didn't cover? Write to us and we'll help you get set up.
Ask us anythingA machine that stops checking in is counted as offline after 90 seconds, and devices move to the next one right after. When you take a machine down on purpose (stop serving on it first), they move right away.
/ Independent · bootstrapped · no investors
Built by a person,
not a fund.
I'm Endi, and I build Underpass myself. No investors to answer to, only the people who run their networks on it. Every message that comes in reaches me directly.