A plan sent every machine running a module at once, ignoring the module's
upgrade policy. Unless the policy says together, the first machine by name
is sent, recorded in the plan, and the rest follow only once its report
after the send says it applied; a failed first machine stops the plan.
The controller's machine moves it from the container to a process by
starting the process first and removing the container once the process
is up (mesh-host's `replaces`). For that moment two controllers share the
store and the bus. Checked what each does:
- the seat's verbs: a queue group per seat, each call answered once. Safe.
- the controller's consumers on CONTROL and EVENTS: push consumers with
no delivery group, so the second bind is refused with "consumer is
already bound" and serve exited. The process would restart for ever,
the host would never see it up, and the container would never go. The
second controller now stands by and binds when the first lets go
(tested on a real bus; fails without the change).
- plans: read, changed and saved whole by the 30s timer, by build
outcomes, by a merge and by `plans stop`. Two timers would each ask a
tier the other had just asked. Working the plans now takes a
session-level advisory lock on the inventory: the timer skips while
another holds it, the other paths wait for it. Build asks happen only
inside plan work and are covered by the same lock.
Builds of one module in flight together finish in any order, and the mesh
took whatever it heard last as what the module is: RegisterModule overwrote
the module's manifest unconditionally, and Held/BuiltAgainst/ReadRepositories
ordered builds by when they were recorded. A postgres build asked before the
mesh-tools runtime fix finished after the one asked after it, and the next
push deployed the stale image (novox/hq issue 219).
A build is now ordered by when it was asked, read from the build-<nanos> id
the controller writes: build.asked and module.built_asked (migration 0055).
A registration from an earlier request than the module's current one is
recorded and refused as superseded. A plan takes as its outcome only a build
asked at or after its own ask, so an earlier plan's leftover build cannot
settle a later plan. Ids of any other shape keep the old order.
A build asked at a commit recorded that commit as the module's ref. Every merge after it failed to
match the module and its plan left it out without a word, and every plan that rebuilt it asked for
the same old commit again. Registration now keeps the branch the module followed (the default
branch for a new one); matching and re-asking read a recorded commit as the default branch, which
heals records already pinned this way; and a merge says which modules of its repository it leaves
out because they follow another branch.
A merge to the controller's own repository replaces the controller in its first tier; the build
that produced the new one was recorded, the plan never heard it, and it waited for ever with every
later plan behind it. The record is the fact: a build recorded after the ask is the tier's outcome,
whoever was listening when it came.
A plan is ordered by artifacts and says nothing about what must be running before what (ADR 0162);
on 2026-10-03 that put the build machine in tier 0 and the controller in tier 1, and the new build
machine could not bind the worker the old controller had defined. One running order enters the
graph, named as its own edge: a module claiming the build seat follows the control plane, and the
built-by edge from the control plane to that holder yields to it — the controller is built by
whichever build machine is running, as the runtime image always was. The edge orders a plan and
never widens it, like built-by.
Given a repository and the files a branch changes (or the modules by name), plans answers with the tiers
the merge handler would produce — the modules those files touch, what packages their source, everything
reachable along the dependency relation — and what each tier does: built and sent to its machines, or
built and left because its policy records. Saved nowhere. On the seat as plans {repository, paths|modules}.
A dependent rebuilt because its base moved, or a module that packages another repository's source,
keeps its source commit; the catalogue announces no move for it, and its machines kept the old image
until somebody pushed — forty-three modules after every runtime-image merge (hq issue 189). When a tier
is built, the plan now sends the machines of every module in it whose policy rolls out, once, moved
commit or not; a module whose policy records is built and left, as its policy says. The gate between
tiers waits as before for the ones a later tier is built by.
A tier whose module a later tier is built by waited for the machines running it to report after the
build — and relied on the catalogue's moved event to send them. A rebuild from the same source commit is
not a move the catalogue announces: the build machine rebuilt for a controller change kept its commit,
nothing sent it, and the plan waited on a report that would never come (2026-10-01, 19:45Z). The plan
now sends the machines running a gated module once, records when, and waits for the reports after that.
The first live plan took seventy-five modules along for a controller change: the builder packages the
controller's source, everything is built by the builder, so everything was reachable. A module built by
the build machine is not changed by a new build machine. Reachability now follows the code and build
edges only; built-by still orders a tier after the build machine and gates it on the machine's roll-out.
plans stop <id> ends a plan by hand: what was asked still builds and registers, nothing further is asked.
The mesh's seat answers plans — the recent plans with their tier, what each waits for and since when,
or one plan whole given its id — so the console reads a merge's progress where it reads everything
else, instead of a person reading the daemon's log.
A module's dependencies are one relation in the catalogue — stands-on, packages, built-by, declared —
answered by one call. A merge takes what moved and everything reachable from it, sorts the set into
tiers (a code dependency in the same tier, a build dependency after its base is built, a runtime
dependency after the build machine is built and running; the build machine's own base comes first,
built by the one that runs), writes the plan to the store, asks the first tier and returns. Every
outcome advances the plan; a ticker advances what outcomes cannot; a controller replaced mid-plan
resumes it. status lists open plans and names one that has waited too long.