FluxOS v9: How Apps Will Be Described Next
Fluxers! Every app on Flux starts life as a small document signed by its owner. It names a container image, the CPU, memory and disk the app needs, how many copies to run and for how long. Flux has no central scheduler: every node reads that document and decides for itself what to run, so every node has to read it in exactly the same way. That makes the application specification something close to consensus, and it is why it changes carefully, one version at a time.
The next version, v9, is the biggest change to how applications are described since the network launched, and one of the headline items on our Q4 2026 roadmap. Here is what it brings.
Where we stand today
Version 8 introduced private enterprise deployments on ArcaneOS. It has become the network’s standard fast. When we published the Flux v9 whitepaper in September, 73% of apps used it. In our November census it was 88%. v9 builds on that foundation.
A cleaner deployment model
- Components as a named map instead of a positional list, so each part of an app is known by its name.
- Plain field names.
repotagbecomesimage,rambecomesmemory,hddbecomesrootFsGb, andcomposebecomescomponents. - Named ports instead of parallel lists that have to line up.
Load balancing you declare
Today, much of how traffic reaches an app is inferred. In v9 you declare it, per route, for HTTP and TCP:
- scheme and HTTPS redirects, custom domains and managed certificates;
- health checks, timeouts, connection limits and retries;
- sticky sessions for apps that need a visitor to stay on one instance;
- backend TLS with a certificate authority of the app’s own, and connection draining, so a node that shuts down lets its open connections finish first.
Draining and backend TLS are already working in the v9 development work. With backend TLS, each node running the app gets a 30-day certificate from the app’s own authority and renews it automatically.
Storage, replicas and placement
- Proper persistent storage, with sized volumes and an explicit map of where each one mounts.
- Replicas and placement as part of the specification, including active-standby and active-active, and a start order that waits until an app’s data is in place. It builds on the primary and standby storage that more than 300 public apps already use today.
- A cleaner encryption envelope for private specifications.
Built for what comes next
v9 also carries the fields that later parts of the roadmap rest on:
- Duration as a first-class field, the foundation for subscriptions, auto-renew and a single balance that warns you before anything expires.
- A payment memo, so a payment names the workload it belongs to. That is the basis for paying node operators from usage under Progressive Node Rewards v2.
- A referral code, with referral rewards paid as hosting credits.
- A machine type, for the VPS tier on the roadmap.
How it will roll out
Every version from 3 to 8 was switched on for the whole network at a set block height. Version 8, for example, arrived at block 1,932,380. v9 follows the same careful path. The specification is published as an RFC for comment before rollout, and it is tested like consensus code, because every node on the network has to agree on it. Apps on earlier versions keep running as they are.
For developers, v9 means an app described the way you already think about it: named parts, plain fields, the traffic rules you want, and storage that says what it holds. For the network, it is the base that subscriptions, operator rewards from usage, referrals and new machine tiers will build on.
Follow the work on the Flux roadmap, and read the full design in the v9 whitepaper.
Posted in Education
by RunonFlux
Tags:
Comments
Leave a Reply
You must be logged in to post a comment.
