pre-alphabuilding in public
High availability without the high drama

Fail over
without falling over.

Service-aware failover, elasticd is an HA daemon designed to move a VIP between eligible Linux nodes using fresh service-health evidence and explicit ownership.

vip/web · ownership playgroundlive state
edge-1owner · healthy
edge-2standby · healthy
edge-3standby · healthy
12:04:29 quorum healthy · 3 voters
12:04:30 edge-1 holds authority for vip/web
12:04:30 route published · 203.0.113.20/32
click here
A small, stubborn gift to operators from fellow operators.
The entire promise

One VIP. One owner.
Everything else may move.

The public contract stays boring. The machinery underneath is allowed to hustle.

01 / public edgeVIP

The address stays put.

Clients keep one destination while ownership moves underneath.

02 / authority1x

One node gets the pen.

Current authority is explicit, leased, and never inferred from vibes.

03 / reconcileLOOP

It keeps checking.

Health expires. State is reconciled. Failover is work, not wishful thinking.

The control loop

Built for the boringly correct outcome.

Four jobs. One daemon. No victory lap. The loop simply starts again.

01

Watch health

Fresh external evidence decides whether a node may contend.

02

Own exactly once

Raft or another configured driver grants current authority.

03

Move the VIP

The authorized node reconciles the local Linux address state.

04

Tell the network

BGP or an external publisher reports the owner upstream.

Failure is the interface

Pull the plug.
We’ll wait.

Infrastructure gets interesting when something stops answering. Pick a failure and watch the desired state change.

scenario/node-gone

The owner disappears

Authority expires before another node safely takes over.

12:04:31 edge-1 authority expired
12:04:32 edge-2 acquired vip/web
12:04:32 route publication reconciled
Open source · built in public

Serious infrastructure. Made by people who enjoy it.

Service-aware failover for active-active services. Every eligible node can stay ready while one explicit owner holds the VIP. Fresh service-health evidence decides when ownership should move. elasticd is pre-alpha—read the design, challenge the failure model, and help turn the architecture into something operators can trust.

Come break it with us