mesh/check being carried to 76c7ddfc5227 on ffa63d25a066 | from a4713e2d28d7
Review round 1: a have naming no build in a phase but running is also written for a
module that runs inside the node's own (waiting for a yes), so taking it out would
drop its resources before the yes. The fp the held have now keeps already fixes
postgres (issue 1289); a test drives an interrupted backup end to end.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
On novox (2026-10-31) postgres's repeat backup was interrupted by the node-engine's
restart: its held have named no build, so the node's own composition took postgres in
and was refused for its secret reader, and nothing ran the backup again. Issue 1289
(R4c, its delivery held).
- A held have names what runs: read from the module's have before backing-up when
the agent has no running entry, as the not-composed mark does.
- A have in any phase but running that names nothing is not run inside the node's own.
- A backup interrupted by the agent's stop is no failure: the next start backs the
module up again and applies while the want asks a backup. A real failure stays
held for the operator's answer.
Quick fix: a new backup ask by hand. Thorough, chosen: the fault fixed where it lives
in the node agent, so every data module's interrupted backup resumes by itself.
Claude-Session: https://claude.ai/code/session_01NcbgfJ9Gum47zkFKWRaovL
mesh/check failed: shanks: the check of novox/mesh-controller#487 failed: go mod download in modules/route-proxy: go: git.novox.be/novox/mesh-controll…
mesh/check being carried to da0f8f1cfdb4 on bb8383b065a1 | from 62b211bb6040
Every put under want.<node>.>, every module's take-in and every placement write
made a pass, and each pass asked the vault for the node's own secrets. The
shadow's change check moves onto the composer, gains the want (by its
fingerprint, the only part composed) and serves both loops.
On the bus node restic's manifest replaced its process measure by awake (mesh-catalog 35ff67bd). The record of restic.measure was the node's, made while restic was folded, so restic's own apply never saw it, and the guard against removing a placed module's resources refused the node's own apply whole every two minutes from 2026-10-30 19:42: nothing of the node's own applied.
Quick fix considered: forget the one record by hand (a hand act on the node; refused). Thorough fix (this): the node's own apply keeps such a record and says so, never refusing or removing it; a module's own apply owns a node record whose id names it, so it alone removes it once neither composition declares it.
mesh/check being carried to b0d6684d3731 on 91fb323ce0cf | from mesh-sdk dec3bfe7a8cb | novox/mesh-host#185 in group fix/compose-retries-transient
The verb compose has no warrant (ADR 0563 §2): Agent.Retry released a module held for a failed backup and re-applied
one already running its want. RetryFailed tries again only a failed apply not held, and refuses in words otherwise.
A hand ask no longer restarts the growing bound, and a pass less than askedWithin old is the answer, so a caller in a
loop drives no passes against the vault. A composer whose loop ended refuses at once. The retry line said try N+1.
The test's first shadow now ends before slowAfter is set, a race the race detector found under the new timing.
On 2026-10-31 ace's route-proxy said only that the vault could not be asked, and ace said nothing for five minutes
while a pass waited after the vault restarted; the retry that already existed was invisible. A refusal for a seat or
a read that did not answer now says when the node composes again ("the node composes again at HH:MM:SS (try N)"), a
retry pass says it starts, a pass that runs on says every minute what it waits on, and the node-engine's verb
compose makes a pass now and says how it went, so a person recovers without a restart.
Four providers waited for a yes on 2026-10-31 and nobody could read what waited: the have said only
"where its data is kept". The read verb composition answers it from the engine: the latest revision of the
want's build or the one asked by digest, every revision the bucket keeps, what runs and where that is known from,
the want and the have, the move's judgement and whether it waits for a yes, and with diff the resources that
differ, a received file naming its consumers. revisionNamed now reads the same revisions.
waitsForYes returned at once for a composition equal to what runs, and settleOne then found nothing to apply, so
the have kept phase waiting-yes for a move that no longer existed (ace.postgres, novox.postgres, novox.keycloak and
novox.mongodb on 2026-10-31). The have is now rewritten to running.
Quick fix: rewrite the have where the equal composition is met. Thorough fix (not here): one place that writes a
module's have from the settled state after every settle, so no early return can leave a stale phase.
The review of d5e8458 found that mesh-wireguard, provided by the mesh and never wanted at a fingerprint, kept the kept
declaration read on every node for ever, against ADR 0550 §2 once the lease is written. A selfcompose row shows the
node's own composes it from delivery.ProvidedManifest with nothing running, so the kept read holds nothing of it.
Quick fix: name mesh-wireguard in allHave. Thorough fix: allHave counts every module delivery.ProvidedByMesh says the
mesh provides as covered. Chose thorough: one list of provided modules, mesh-sdk's.
At 19:43:01 on 2026-10-30 shanks' new engine said "the kept declaration is no longer read: each of the 77 modules
placed here has a have of its own", and from 19:43:57 every apply of the node's own was refused: it would remove
systemd-resolved's resources (novox the same for nats and restic). systemd-resolved's have names only a controller
build (build-1791404191702719760), no fingerprint, so the assigner wants it nowhere and it runs folded in the node's
own, of which the kept declaration is the one record; retired, the folded module was composed as nothing.
Quick fix: never retire the kept declaration. Thorough fix: allHave counts only a have naming a fingerprint of the new
form (or removed); a module at a build of the old form keeps the read, named with why, until it has one. Chose
thorough: the read retires by itself once every placed module is wanted.
mesh/check checking 33131dcbfe77 on 2ae7c53376e8 | novox/mesh-host#181 in group fix/build-not-done-holds-one-module
At 19:30 on 2026-10-30 node-tools' want moved to a build not yet done; once it landed, ace, g14 and shanks stayed
refused until some other input moved (g14 and shanks at 19:40, ace after a restart), since the composer watched no
build.
Quick fix: compose again on a timer while anything is held. Thorough fix: the composer watches the builder's builds
(within the node's compose reads, NodeComposeReads) and makes a pass when a done build of a module the node wants, at
its want's or the trunk's build, lands; nothing else moves it. With mesh-sdk 9685990a (pinned here) a running module
whose build is not done is held, keeping what runs, so the pass that waited composes it as soon as it lands. Chose
thorough: event-driven, no poll.
mesh/check checking 4055e0094b48 on ee71ba4dea81 | novox/mesh-host#180 in group fix/one-bad-item-holds-one-module
At 19:13 on 2026-10-30 power's bar item (mesh-catalog ca8d6f33) made frombus.Compose refuse the whole node, and the
composer marked every module not composed (32 of 81 failed on shanks). The composer already holds what Compose names
held (NotComposedEach, issue 1126); the SDK now names a refused manifest's module held instead of refusing the node.
Quick fix: none in mesh-host; thorough fix is the SDK's input (one rule for Compose, Answer and PlansHeld). The test
replays the captured power manifest and fails on the old pin with the live words.
mesh/check checking b9c24a7b42cb on 737bce9c9448 | novox/mesh-host#179 in group fix/sweep-2026-10-30
mesh-vault (mesh-catalog f3b13de2) answers a move it refuses with `refused: {reason, held}` and `why`, and
no read-back. The engine read only the read-back, so every such row failed "answered with no read-back"
and the controller asked it again on every pass.
Quick fix: treat any answer without a read-back as done. Thorough fix (chosen): read `refused`. For
holds-another-value the vault's value is the live one: the move answers done, superseded, at the held
generation with the held record in its words, so the controller records the row and stops asking; the
journal says it once per secret. Any other reason fails in the vault's own words.
restic's composition dropped its process restic.measure on g14 and shanks at 16:14; the engine removed it
and said it down on every look after. The judges were set only when facts were said, and the facts said
inside an apply read the module's composition from before it (the agent records the new one only after
the apply returns), so the removed resource was put back. Judge.Set itself prunes what is not declared.
Quick fix: set the judges after the restic apply. Thorough fix (chosen): one function hands every judge
(liveness, resolver, units, accounts, libraries) what is declared; it runs from the facts as before and
now also whenever a module settles (applied or removed) or is taken as running, so every removed health
entry of every module is pruned at recompose.
On 2026-10-30 ace's redis, postgres and restic (14:03), step-ca (13:55) and mesh-vault on novox (15:22)
were held on a single backup ask that met no responder while the tool runner restarted.
Quick fix: ask node-backup.now twice. Thorough fix (chosen): every node-backup ask (backed-up and now)
is asked again with the engine's backoff while it is known never to have been delivered (ErrNotAsked),
bounded at three minutes; an ask that may have been delivered (a timeout, a refusal) is never asked
twice, so no backup starts twice. Past the bound the backup fails saying how often and how long it was
asked, the module is held and its have carries those words for the assigner's ask and condition. This
was the engine's only one-shot seat ask.
Every engine's node-seat subscriptions were refused at 12:10 by a user list composed without them. nats
asks a refused subscription again only on a new connection, and a user-list reload makes none, so the
seat answered nobody all afternoon and the messenger held every Telegram ask. The endpoints also lacked
the kind and scope discovery reads, so the console said nothing in the mesh is called node-engine.
Quick fix: restart the engines. Thorough fix (chosen, the second time a stale user list left an engine
deaf): a refused subscription of any kind is said and the link is opened again after a pause (1 min,
doubling to 30 min), so every subscription the engine makes is asked again; the seat's endpoints carry
kind, seat, scope, module, description and schema like every tool runner's.
mesh/check checking 015334020ad2 on a3c098a5c618 | novox/mesh-host#177 in group feat/774-l3-reland
The conflicts were go.mod, go.sum and testdata/beside/CAPTURED alone, taken from the release. 2da0de7 is 6e4a301's code;
its conformance events are unchanged.
mesh/check pass 2fd165f5af8e on 8c5af3e4a3a3 | unit 9174 run, 546 kept, 51 skipped | integration 9952 run, 659 kept, 60 skipped | over bound: unit, in…
78c9ec8 is sdk main 307b623, whose tree is main's pin dbcaa29, with one commit that parses the new words (a credential's
form, an artifact's bus, a log's max-age-days) and names the new seats; an offer with no form still means pair. No code
of ours changes. Unit and integration suites pass against a local bus and main's catalogue and controller beside: 1260
tests, 7 skipped for a container or a throwaway machine.
No code of ours changes. Whole unit and integration suites pass at main's code against a local bus: 1260 tests, 7
skipped for a container or a throwaway machine.
The whole integration suite passes at this head against a local bus (JetStream, and a bus with the witness grants) and
the release's mesh-catalog and mesh-controller beside it: 1260 tests, 7 skipped for a container or a throwaway machine.
Every account resource marked root never goes into NodeHealth.Root; the statement's contract into Contract. Relogin,
root and found are judged each on its own, so an account may be both. Pins mesh-sdk f3677eb. Test: the report's row in
TestAHealthStatementIsSaidOnTheLinkOpenNow carries the agent account's verdict and the contract.
delivery.NetworkPart.Targets (mesh-sdk 6eae076) is filled where the report is made, by the controller's own rule:
- the hub's node for "hub";
- each resolver's node by its address on the private network;
- none when the part points at nothing, at this machine, or at an address no other mesh machine has.
The names come from the placements the node's own composer holds (Decided.Address, Decided.Hub). A spoke's tunnel names
only the hub, so the tunnel cannot name a resolver on another spoke. A node that does not compose its own names no
target.
Test: rows in TestAHealthStatementIsSaidOnTheLinkOpenNow (hub, two resolvers, a VPN's resolver, its own, a healthy
part).
delivery.NodeHealth and its parts replace this package's copies. The message's network and units parts are aliases of
the same types, so its JSON is unchanged. The engine keeps the derivation from its statement and the write on a change.
Test: the key's bytes, decoded and encoded again by the sdk, are the same bytes.
Until now the six machine alerts came from a health message passing to the controller. The report is now kept in the
engine's declared state `health`, key `<node>`, created by the grants part as `acts` is. Only this node's engine
writes it, after comparing it on every look and only when what it says changed; it is read first after a restart.
It carries:
- the network parts (machine-network; network-rewritten through the resolver file part's writer and owner);
- the tunnel's peers, each seen within five minutes or not (network-unreachable);
- the failed units no module owns (machine-units);
- the accounts marked relogin needed (relogin-needed);
- the directories used as found (directory-used-as-found).
The streaks and the time of the look are left out, so a look alone writes nothing. A peer is reported seen or not,
never its handshake time, so a renewed handshake writes nothing.
The layout is internal/link's (NodeHealth) until it moves into mesh-sdk. Pins mesh-sdk 58a8324.
Tests:
- rows in TestOnTheHubAnyFreshPeerIsAHealthyTunnel: the peers, and a change only when one is seen or lost;
- a case in TestAHealthStatementIsSaidOnTheLinkOpenNow: what the report takes, and one write per change across a
restart;
- the manifest's states in TestTheManifestConsumesTheBackupEvent.
- A placeholder is `${secret:<name>}`, the name the module calls the credential (catalogue.SecretLocal): the local
name where it keeps several values of one provision, else the provision. The `<provision>.<local>` form goes.
- The engine finds the one key under `<node>.<module>.` of mesh-provisioner's `provisions` going by that name. That is
what node.<node> may read, by a watch to its caught-up mark; plans are not read.
- The reference's ID must be a credential name of that provision (provisions.VaultName). The vault's answer must carry
exactly that name and generation. Anything else refuses the write, naming both.
Pins mesh-sdk 66ff33d. Tests: rows in TestAWriteAndAnApplyOfTheSameFileSerialise (a local name, a name no provision
goes by, a reference to no credential of the provision, a mismatched answer).
mesh-tools
jschoubben