A named push flushed every other machine whose declaration differed from
what it was last sent (hq ADR 0083). Under an upgrade policy of `record`,
or a plan still waiting on its first machine (ADR 0218), every machine
running the module differs, so `push <one>` sent the held build to all of
them (hq issue 259).
Each send now records which build of each module it carried
(node.sent_builds, migration 0061). The cascade, and the bus holder added
to a named push, skip a machine any of whose modules would move to a
build its policy records or an open plan has not sent it, and say which
module, which build, why, and that `push <node>` sends it. A machine
whose last send was not recorded is held until it is named. The named
machine itself, a whole-mesh push and `push --behind` are unchanged.
A registered replay asked now outranked newer asks of its module, and a plan
took any later outcome as its answer. A replay is now refused while the module
is asked anywhere, a plan module asked under an id is answered by that id
alone, and a rebuild of a commit asks what the module follows. An ask handed
back after a restart no longer reads as dead; a cancel that meets a start is
withdrawn; pause holds for an ask fetched as it lands; kill removes containers
before and after the build ends and says whether its outcome went out; a
holder may say only its own machine is paused.
The holder's verbs were served and found by nothing; it now answers discovery
with its machine's four, as a runtime announces a seat's verb, so the console
finds <node>/node-build-agent.kill. A cancel publishes no outcome, so a build
waited for also looks at the cancelled set. A plan that stopped at its first
machine is retried by sending that machine the module again, unless a newer
plan holds it.
Nothing showed what waited for a build machine, and an ask could not be
dropped without leaving the plan that made it waiting for ever. New verbs:
queue, cancel, clear, rebuild, replay, kill, pause, resume, and plans retry.
Every ask a person drops is recorded failed through the same take-in as a
failed build; a plan keeps the id it asked each module under and matches
its outcome by it. replay is a dry run unless registered, and registering
an older commit than one registered since needs --older (hq issue 207).
A plan waiting on a seat paused on every holder says so and is not late;
a failed plan can be retried, and a rebuild joins the plan holding the
module instead of running beside it.
A queued ask could only be waited out and a running build only ended by
stopping the machine, which redelivered it elsewhere. The holder now serves
current, kill, pause and resume on its own machine's subjects; a kill ends
the build's process group and labelled containers and settles the ask as
failed, killed by hand; pause is kept in the workspace across a restart and
said on the bus. The controller writes cancelled ids to a cancelled set the
holder reads on taking an ask, closing the race a delete alone leaves.
With one machine first a module is sent twice, and the tier gate asked every
machine for a report after the second send: the first machine's report, made
between the two, read as stale and the plan waited for ever (hq issue 256).
Also round a plan's wait to the second, not the minute, so it is not 0s.
A first machine that does not report within the bound now stops the
module's rollout, naming it. The first machine is the first by name heard
from lately; reports are judged by what the store says was last sent. A
plan that ends says what it built and never sent, and a failed send's
error is kept in the plan's note.
Grants first could hold back the very declaration that lets the controller
issue them. The holder now goes first, then buckets and memberships, then
the rest; a membership that fails holds back only its own machine, and an
announcement whose send stopped at its grants is asked again. Whether the
holder must go first is read from a digest of the user list it was last
sent, not its whole declaration. Migration renumbered to 0058.
The catalogue's upgrade announcement sent every machine one after another
without waiting for any to apply, beside the plan that now sends one
machine first. A module an open plan has not finished sending is the
plan's to roll out.
A merge planned without looking at open plans, so two plans worked the
same modules and a stuck plan stayed open for ever. The newer plan folds
in what older plans of the same repository and branch had not built or
sent, and closes them as superseded. A person can close a stuck plan by
id with `plans close <id>`.
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.
A module's new state reached every machine before the permissions to use
it, which came only with a later push. Memberships and buckets now go
before declarations, the machine holding mesh-broker goes first when its
user list must change, and a grant that fails sends nothing and is an
error so the rollout is retried.
A merge adding a module the mesh has not registered rebuilt every module
built from the repository. A path under a directory known to hold modules
belongs to that module when its module.json is among the changed files.
The store's garbage-collect marks only from manifests, and archives were
published as bare blobs, so the first real collection would delete every
archive the mesh keeps. PublishArchive now puts a deterministic OCI holder
manifest (empty config, one layer) beside each archive; the sweep holds every
kept archive before it lets anything go, which backfills existing bare blobs,
and lets go of an archive holder-first. A forgotten module no longer keeps its
five recent builds (ADR 0189). `collection [--json]` reports kept archives
held/unheld and what may be let go, so the dry run can be lifted on evidence.
A consumer made with the server's default replays everything a stream that
keeps history holds: the controller's EVENTS consumer, re-made that way,
replayed a week of merges and builds one at a time and held every new one
behind them. FromNow makes it start at the end; broker consumer-reset
re-makes a stuck one from now, refusing a work queue. hq issue 244.
A dry run of an unreviewed branch was heard by the daemon like any build, registered, and its
definition reached a machine (novox/hq issue 240). The mark now travels with the build and the
daemon records, registers and plans nothing for it.
Every assigned module was composed as a bus user, though only one declaring
a broker secret can ever be issued an account; the rest were named on every
status, plan and push as credentials never minted (137 now), burying the
real gaps. Their durable consumers are now derived from what the runtime
carries, so nothing they hear changes. The composed file is unchanged:
those users had no password and were already left out. Fixes hq issue 195.
- mesh-dns-resolver: a mesh seat delivering wildcard-resolution, so every node's resolver
configuration resolves to its one holder; node-dns-resolver kept until nothing claims it.
- ${bound:<provision>:address}: the providing machine's private address, for the one consumer
that cannot use a name — a machine's resolver configuration.
- zone: a module declares the zone it answers and the listen that answers it; the controller
settles it per node, refuses duplicates and shadowing, and hands the resolver .Zones to forward.
- node-hosts-file: a node seat whose holder owns /etc/hosts, with entries/add/remove.
The resolver tests follow the catalogue: no runtime dns (containers copy the machine's resolvers),
live-restore held by resolv-conf, resolv.conf naming the resolver by address then a public one.
The environment and shell contributions were written nowhere on a node without their holder;
they now derive a dependency on node-environment, node-login-shell or node-display-server, met
and refused as ADR 0207's are. Two modules declaring one package, path or unit made the node
unresolvable after the assignment was recorded; that is refused first now, because no later
assignment can complete it.
status reported no unmet dependency on any node once systemd, pacman and docker
were assigned to all four (to-be 42), which is the condition ADR 0207 set for the
switch. A dependency no catalogue module could meet stays a report before and
after the switch, as assign already said it: there is no remedy to name.
assign and unassign say only what they changed on their node; push <node>
lists that node's, push to many counts each and points at status. The
once-per-change log is the serving controller's alone: a one-shot command
starts with no memory, so it logged every node on every call.
Seed node-package-manager and node-container-runtime. Derive each module's
dependencies from its declared service, package and container resources;
judge them over the node's whole set, exempting the foundation. Refuse at
assign (several modules may go on as one act) and at unassign of the last
holder; report at composition in status, behind one switch.
Issue 225. The mesh seals one credential per consumer beside the provider's
contributions file, and wrote it root-owned. That was right while a module's
own code ran in a container as root; ADR 0198 moved that code under the node's
runtime, as the node's account, and the secret stayed root's. On the control
machine two consumers went unprovisioned for three hours and the only sign
was a line reading 'secret not readable yet', 4330 times.
The same sentence is already written for a module's own secrets a few hundred
lines above — 'a root-owned 0600 file is one that process cannot read'. This
is that rule reaching the other kind of secret the mesh writes for a module.
Issue 226. The sweep met a reference recorded with the store's old address,
read 'I will not address this' as 'the store refuses everything', and
collected none of the 1681 it had found. Two changes: references from build
records are read through Recorded, where the provenance is known — not in
LetGo, which cannot tell one registry host from another and must stay strict
— and a reference the sweep will not address is now ErrNotOurs, skipped,
never a reason to stop. Only the store refusing ends a sweep.
make check: the two failures both fail on main as well — the resolver test
(hq 202/203) and the service-manager test, which reads this machine's own
shell environment.
The raise at start was the only place buckets were asserted, so a module
registered and assigned since had none until the control plane restarted —
found on the first module to declare state.
Three things found reading this back, each of which would have been quiet.
A consumer that keeps several holders of one provision (ADR 0094) gets a
login per holder, and a provider derives from the login — so it would make a
resource per holder while the consumer is told one value for the requirement.
That is issue 124's own failure one case to the side: authenticate, then be
refused on every object. Refused now, naming both ends.
The sweep runs inside somebody's build and was unbounded. At most two hundred
artifacts and sixty seconds, stopping at the first refusal because a store
that refuses one refuses all; the rest is offered again next build.
The citation and migration renumbers are in the commit before this one.
A manifest names the state it keeps (state) and reads (reads); the controller
asserts a key-value bucket per name on every raise, grants owners write and
readers read (measured against a running server), issues each assignment its
buckets in the membership, and reports buckets nothing declares without
removing them.
The mesh names what may go from its own build records — a digest it did not
record making is never named, which is what keeps the sweep away from the
images genesis pushed. An artifact stays because a definition the mesh holds
names it, or because it belongs to one of the five most recent successful
builds of its module.
internal/artifacts asks the store to let go of one; internal/inventory
decides and remembers (migration 0055); the sweep runs after a build the mesh
recorded, which is when both the bytes and the keep set moved. Never fatal to
a build.
And the manifest side of while-stopped, refused from the definition alone:
no schedule, run-once, a container the module does not declare, itself.
The controller is a Go program and was the one piece of the mesh's own Go
code still shipped and run as an image (novox/hq issue 213; ADR 0188 §1:
a module's own code is bundles; §3: a service bundle is a process).
The manifest now builds one Go bundle, `controller`, and runs it as the
process `mesh-controller` (`./mesh-controller serve`) under an account
the module declares. What the container gave it, replaced:
- host network: a process is on the host's network; nothing it reads
names a container network
- user 65534: the account `mesh-controller`, which owns its secrets and
its state directory
- the eight mounts: the env names the host paths the mesh already places
(the store, broker and bus files under the state directory, the
broker's certificate under /var/lib/mesh-broker-tls); the `broker`
mount was read by nothing and is gone with the others
- `container-runtime` is no longer required on its machine
Its preparation is the same binary with `prepare`, as a run-once process,
and the process `replaces` the container `server`: the host keeps the
container answering until the process is running (mesh-host). Needs the
previous commit live in the running controller, and the host's
`replaces` on the controller's machine, before it is registered.
No image is built by the mesh any more. The Dockerfile stays for genesis
and the lab (`make image`, its Go base now pinned in the Makefile).
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 manifest names its toolchain by language, not in build.on, so the planner did not know a bundle
depends on the module that publishes its toolchain and built the two in one tier: the bundle
against the old toolchain, recorded as built from the new commit. The edge is read from the
manifest, so it holds before any build recorded it, and a toolchain that moves rebuilds every
bundle compiled in it.
Grants: a principal that serves tools subscribes $SRV.PING/$SRV.INFO and those questions under
each name it serves — its own and no other's; the tool runtime and people may ask. The controller
answers discovery for the mesh-controller seat in NATS's services format, one endpoint per verb it
serves, with the seat's description and schema. module list --json says which modules declare tools,
so the console expects an announcement only from those.
The console's discovery reads the machines and the modules; parsing a printed column breaks when it
is reworded. Both now answer JSON on --json, as status and seats do, and the seat verbs ask for it.
The roster published routed names — public ones first, then (in this PR's first take) internal ones
told apart by suffix. Neither is needed: a node has one internal domain and every route on it is a
name under it, answered by the resolver's per-node wildcard; a node's public domains are public
DNS's. routeNamesInTheMesh and NamesServed are removed, and a test pins .Names to the machines.
NamesServed read a route's public `name` and plan.go then filtered by suffix — telling the mesh's
names from public ones by their spelling, when the mesh composed both itself. It now publishes the
`internal-name` it composed under the serving node (ADR 0151); the suffix filter is gone.
Every routed public name was published into each machine's hosts region at its serving node's
private address. ace's resolver also answers its LAN, so a phone there got the control-node's
tunnel address for the mail server and could not connect. Routes have internal names under the
serving node (ADR 0151), so only names under the mesh suffix are published now.
jochen
jschoubben
mesh-admin