They came from `requires: secret`, minted by the vault — right for a fresh
setup (the image's INIT_* variables read them once), wrong for an instance
that already exists: setup is skipped, the minted values match nothing,
and the provisioner holds a token the server never issued (issue 100).
InfluxDB will not take a chosen token value, so the operator token must be
accepted from the instance (`secret accept ace influxdb admin-token`); the
password can be either. As own secrets both are minted for a fresh install
exactly as before, and accepted where the data already knows them.
Found migrating ace's influxdb.
A program that checks its resolver validates — Mailu's admin does, at
start — could not use the mesh's resolver, and the one it ships instead
knows no mesh name (novox/hq issue 171). proxy-dnssec copies the AD bit
from 1.1.1.1 and 8.8.8.8, both of which validate.
admin is the one container here that reaches something by a mesh name:
the database, bound in database.env. A mesh name is answered by the
machine's resolver, which the runtime hands every container that does
not name its own (novox/hq ADR 0148); Mailu's unbound knows no mesh
name, and since the mesh stopped copying names into containers admin
could not find its database. The others keep unbound — rspamd needs a
validating resolver for DNSBL lookups and asks for no mesh name.
dnsmasq admits a query by the interface it arrives on when told interface=,
and by the address it is sent to when told listen-address=. A container's
query is sent to the machine's private address but arrives on docker0, so
interface=mesh0 dropped it silently on every machine (novox/hq issue 110).
Name the address, not the interface.
novox/hq 04-ISSUES/110. The module writes the runtime's `dns` key and
deliberately ordered no restart, because a restart stops every container.
It also ordered no reload, and the key holds only for containers created
after the runtime next starts. On two of four machines the runtime
predated the file — one since August — so every container there was
handed a public resolver and no mesh name resolved, while everything read
as fine. The third machine works only because its runtime happened to
restart later.
The file now also sets live-restore, which the runtime reads on a reload,
and the module declares the runtime reloaded when the file changes (ADR
0102: reload, don't restart). A reload still does not make `dns` take
effect; what it does is make the one restart that key needs keep every
container running. That restart stays the operator's, once per machine,
and is harmless from the second time on.
The mesh restarts nothing here. Undeclared, the runtime's unit goes back
to the state it was found in (ADR 0118).
MESH_PROVISION_POSTGRES named 127.0.0.1:5432 literally; the seat twin
(${seat:mesh-store:5432}) corrected it only on the machine holding the
mesh-store seat. On any other machine — ace, where the module provides
postgres-database without the seat — the twin is empty and the sidecar
would have dialled whatever else holds 5432 (HAL's postgres). ${port:5432}
is the machine port on every node; on novox the twin still says the same
6852, so nothing moves there.
mqtt-topic served nothing: with two listens the mesh could not say which port a consumer dials, so a
consumer had to type 1883 into its config. It now serves the MQTT listener's port (the machine's,
once assigned) and the scheme, so `${bound:mqtt-topic:port}` fills.
The provisioner confined every consumer to `<as>/#`, which leaves nothing for the consumers the
broker exists for: Home Assistant discovers under homeassistant/# and tasmota/discovery/#, and
Node-RED's flows follow the devices' own topics. A consumer now contributes `topics` (MQTT topic
filters) to its mqtt-topic requirement and is granted exactly those; with none, its own subtree as
before. Settings merge into contributions, so an operator narrows a grant per assignment. The role
is brought to exactly the wanted ACLs (stale ones removed), `holds` checks the ACLs too, and an
invalid list is refused, never quietly narrowed. Only the role named for the consumer is touched:
a client carried from the predecessor's password file keeps its own.
grafana's data source and Node-RED's influxdb nodes reached ace's
InfluxDB by a LAN IP or a public name nobody routes, with a credential
somebody made by hand. Now a consumer requires influxdb-api and is told
where it is, which org and default bucket it serves, and signs in with
the password the mesh minted for the pair.
The credential is a v1-compatibility authorization, made per grant by
the new provisioner: InfluxDB 2.x generates API tokens itself and
ignores one the caller sends, so a v2 token could only be accepted by
hand per pair; a v1 authorization takes a caller-chosen password (8-72
characters, the mesh mints 40) and reads/writes every bucket as a
database of its name over InfluxQL and line protocol. A consumer
contributes `access` (read, write, read-write) and, for writing, the
buckets; a missing bucket is made and never deleted. Only
authorizations named mesh_* and marked [mesh] are ever changed or
removed; anything else of that name is refused and left alone.
The org and default bucket are served facts the assignment's settings
set, reaching both the consumers and the provisioner's config.json.
The module named /var/lib/letta, a layout no definition may carry (ADR
0112); state is now a placed directory holding the bindings and secrets.
letta had no route, while the server it replaces is reached by its public
name (a workflow calls it there). It now contributes one for its `web`
endpoint; reach is the assignment's.
The server needs an OpenAI key for agents on OpenAI models, and nothing
gave it one: `openai-api-key` is an own-secret, accepted from the
operator (a key someone chose, not one the mesh can mint). The
server-password is accepted the same way where a server already has
clients. Both, and the database password inside LETTA_PG_URI, stay in the
server's environment: letta 0.6.x reads settings from the environment
only, and its startup.sh starts an embedded PostgreSQL unless
LETTA_PG_URI is set - the declared reason now says so.
The runtime's tools never authenticated: the client sent only a Bearer
token, and 0.6.x's --secure mode checks X-BARE-PASSWORD ("password <it>")
and answers 401 otherwise. The client now sends both. Its password comes
from the runtime config file (the key client.ts reads first) instead of an
env-file, so the runtime container no longer carries a secret in its
environment.
Image: the same 0.6.8 image, now pinned by the index digest ace runs
rather than its amd64 manifest.
Verified: catalogue tests with MESH_CATALOGUE set; tsc -p tsconfig.json in
the mesh-tools build image. Throwaway containers: a fresh 0.6.8 with its
embedded PG and two blocks made through the API; stopped, copied, dumped
from the copy; restored (schema letta + vector pre-made by the superuser,
--no-owner --role, search_path set on the database as the original had
it) into a grant-shaped database on the postgres module's pgvector image
(PG17); started in this shape: alembic finds nothing to do, both blocks
are there, a wrong password is refused with 401, and the patched client
lists agents. Test containers and data removed.
The manifest stated /services/redis/data and /var/lib/redis-module - novox's
old layout, paths no definition may carry (ADR 0112). State is now the
assignment's root, grants and data are placed, and the config file, the
secret file, receives and grants all name them as ${dir:...}. Paths inside
the sidecar are its own view and are unchanged.
The data directory and config are owned 999:1000: the image's redis user is
uid 999 in gid 1000 (checked in both builds), which is who owns ace's data
today; 999:999 named a group the image does not use.
Image pinned to the 7.4.11-alpine build ace runs (2026-09-17); the old pin was
the same version, built in August. Older-than-running is never the pin.
Nothing is assigned it anywhere today, so no machine changes.
Verified: catalogue tests pass with MESH_CATALOGUE on this tree; the
declaration composes for ace with every path under /var/lib/redis. The pinned
image ran as a throwaway with a 0600 999:1000 config and a 0700 data dir:
unauthenticated PING is refused (NOAUTH), authenticated SET/GET works,
appendonly is on, the server runs as redis.
The manifest stated /var/lib/mssql, its grants and its SA file by path, and
declared it listens on 4848 - the port one machine's predecessor published,
which is an assignment's fact (ADR 0112, 0138). ace is moving its own
SQL Server (80 GB of work databases) onto the mesh, so the module has to be
the same on every machine.
- state is the assignment's root (place "."), grants is placed, and every
reference (sa.env, the env-file, the SA and grants mounts, receives,
grants, own-secrets) names them as ${dir:...}.
- the database endpoint listens on 1433, the port SQL Server uses; a machine
that must keep an older number pins it in its assignment.
- the image pin is unchanged: it is the digest ace runs today (CU27,
16.0.4295), the same as novox.
novox is untouched: rendered with novox's own setting ({"ports":{"1433":4848}})
through the controller's Declaration and Rules, every resource - paths,
container names, volumes, env-file, the 4848:1433 mapping, owners, modes - is
byte-identical to what main renders; the only difference is the firewall
rule's comment text (still port 4848, from the mesh).
Verified: catalogue tests pass with MESH_CATALOGUE on this tree. The pinned
image ran as a throwaway on a 0700 10001:0 data dir with a root-owned 0600
env-file (dummy SA), answered sqlcmd as sa; a scratch database stopped,
copied with cp -a, checksummed and started on the copy kept its rows and
CHECKSUM_AGG.
The manifest named /services/influxdb and /var/lib/influxdb-module — one
machine's paths — and passed the admin password and token through the
environment. ace is moving its 2022 instance onto the mesh, so the module
has to be what it is on any machine.
- data, config and state are placed directories; the data keeps 1000:1000,
the image's influxdb user, which is who owns ace's data today.
- the init secrets reach the image through its own
DOCKER_INFLUXDB_INIT_{PASSWORD,ADMIN_TOKEN}_FILE; the vault's files are
mounted read-only. secrets-in-environment is gone.
- the sidecar reads its token from the same file (MESH_INFLUXDB_TOKEN_FILE,
added to client.ts) and reaches the server at its assigned machine port
(${port:8086}) instead of assuming 8086. The unused config-dir mount,
which held the CLI's copy of the admin token, is dropped.
- the api endpoint contributes a route: the web UI is how people use it,
and reach is the assignment's to say.
Verified: catalogue tests pass with MESH_CATALOGUE pointed at this tree.
The pinned 2.9.1 image, run on a scratch copy of ace's 2.4.0 data, opens
it, runs its metadata migrations (backing up the pre-upgrade bolt/sqlite)
and hashes the two stored tokens; /health passes. A fresh setup through
the _FILE variables, with dummy secrets as root-owned 0600 files, accepts
the token (200 on /api/v2/buckets) and the password (204 on /signin).
client.ts typechecks strict and reads the token file, tolerating the
endpoints key in its config.
The module stated /services/unifi/data and fixed machine ports (8443:8443 and
eight more) — one installation's layout and numbers, which a definition may not
carry (ADR 0038, 0112). Data is now a placed directory (${dir:data}, 1000:1000,
0700), the container publishes its own ports and the mesh assigns the machine
side; an assignment pins them where devices already know them. The L2 endpoint
names the port the software uses (1900), not the one a machine published it on.
The sidecar dialled https://127.0.0.1:8443, true only while the machine port
equals the container's; it now asks for ${port:8443}. Its controller password
was a setting (plaintext in the mesh DB); it is now an own-secret written into
the one mergeable file (ADR 0086). The username stays a setting.
The web UI is contributed as a route to the "web" endpoint over https with
insecure upstream (the controller's own self-signed tls), as mailu's web-tls —
what HAL's hand-written traefik file for unifi does today.
Image pinned to the manifest list ace runs (8.0.24-ls221); the old pin was its
amd64 child, so the image is unchanged.
Verified: catalogue tests pass with MESH_CATALOGUE set; a throwaway container
of the pinned image on a fresh 1000:1000/0700 data dir answers /status (8.0.24,
up) and /inform; the sidecar client built from a config.json carrying site,
password, username and an endpoints key reaches it and is refused only on the
dummy credentials.
The image seds ICECAST_*_PASSWORD from the environment into /etc/icecast.xml;
ADR 0086 wants secrets as files. icecast starts as root, reads its config, then
drops to uid 100, so a root-owned 0600 icecast.xml rendered with ${secret:...}
and mounted read-only works and the entrypoint's seds never fire (no env set).
The "secrets-in-environment" exemption and server.env are gone.
Also: directories are placed (state, logs owned 100:101 so the image's VOLUME
/var/log/icecast is not an anonymous volume per container, as HAL learned);
the server and sidecar share a module network, so the sidecar reaches
http://icecast:8000 instead of assuming machine port 8000 on the host; the
stream endpoint is routed (label "icecast"), as HAL served it via traefik.
Secrets remain mesh-vault grants (requires secret), now under ${dir:state}.
Verified: catalogue tests with MESH_CATALOGUE pointed at this tree; a
throwaway container of the pinned digest (the one ace runs) with the rendered
file (dummy secrets, root 0600, :ro): runs as icecast, status-json 200,
admin 401 without / 200 with the admin secret, a source PUT with the source
secret mounts, a listener receives it, a wrong source password gets 401, logs
land in the uid-100 directory.
The module stated /var/lib/mosquitto-module and /services/mosquitto/data —
novox's layout, a path no definition may carry (ADR 0112). State, grants and
data are now placed directories (${dir:state}, ${dir:grants}, ${dir:data}),
the admin secret lives beside the broker account under the mesh's own state,
and the receives/grants maps follow the grants directory. Paths inside the
sidecar are its own view and are unchanged.
Image pinned to the 2.1.2 build ace's predecessor runs (2026-09-17); the old
pin was the same version, built in June.
Found preparing ace, whose broker carries a password-file user (an IoT switch
and home-assistant). Carrying it is a data step, not a manifest one: the
migration repo has scripts/mosquitto-pwdfile-to-dynsec.py, which moves $7$
PBKDF2 entries into the dynsec store hash-for-hash (tested end to end).
The route binding still named /var/lib/searxng-module, the directory the
previous commit placed elsewhere — the host would have written it into a
directory nothing declares. Same shape as gitea and nextcloud.
The module ran searxng on the image's built-in settings, which serve html
only — so the module's own search tool (format=json) was refused by the
software it fronts. And there was no way to configure it per machine: the
only file settings reach was the sidecar's.
settings.yml is now the module's one mergeable file (JSON is YAML): generic
defaults in the manifest (json format on, limiter and image proxy off,
valkey wired), and whatever differs per machine — base_url, method,
autocomplete, suspended times — set as the assignment's settings. The
secret key is filled on the machine through ${secret:secret}, so the
secrets-in-environment exception and the env file go. Directories are
placed. Image pinned to 2026.9.20, what ace runs today (the old pin was
older, 2026.9.1).
The sidecar's config.json is no longer mergeable: settings merge into every
mergeable file of a module, and the sidecar would have received searxng's
keys. It only ever read an optional url, which its env already carries.
Verified on ace: the pinned image serves html and json from a read-only,
root-owned 0600 JSON settings.yml.
novox/hq ADR 0147, issue 129. Every internal HTTPS name fails verification
on every machine: the certificates are genuine and nothing on a machine has
ever been told what issued them. The proxy's fetch answers for the proxy and
for nothing else — a browser, git over HTTPS and every module calling another
by an internal name read the machine's own trust store.
The module requires internal-acme-ca, fetches the root over the mesh's own
network (no prior trust to have; that is what this establishes), installs it
among the machine's anchors and refreshes the extracted bundles. Being
unassigned stops the unit, and stopping it takes the anchor away and
refreshes them again.
Arch's layout is named out loud: a machine that keeps anchors elsewhere fails
visibly rather than writing a file nothing reads.
What was in the catalogue was the first thing I built, not the thing ADR 0146
describes. It dialled raw ports on machine addresses from one hosting form and
emitted nothing, so findings would have sat in a file on the machine — the exact
thing issue 145 is about. It was never registered, never assigned, and never ran.
0146 says names per hosting form, fetched over TLS with the certificate verified,
and machines discovered over the bus. That shares nothing with this but the word
checker, so it goes rather than being bent into shape. Recorded as work to be
analysed and built deliberately.
Connectivity is checked by hand in the meantime, against the services the mesh
already runs.
The mesh asserts three things are callable (ADR 0144) — what runs on the same
machine, another machine's service exposed to the private network, and another
machine's service exposed publicly — and has never checked any of them. The first
was broken for eleven hours while the mesh reported every machine healthy.
This runs on every machine, on the cadence the mesh already has, in its own
container: the same position every other module calls from. Not the host and not
the control plane, both of which reach these addresses by paths no ordinary caller
uses and would have passed throughout that outage.
**Its probe is its own endpoint, and that is the point.** Declared reachable over
the private network like any other service, so it is admitted by exactly the rule
that governs every internally-exposed service and fails when that rule is wrong.
The tempting target is a service every machine has, and those are the ones never
closed — ssh above all — which would have passed while the thing that actually
broke was a service exposed to the private network.
It resolves before it dials and says which failed, because a name that does not
resolve and a port that does not answer have different owners. One failure is not
a fault: a machine rebooting is ordinary, so a path is broken after consecutive
runs and the count travels with the result. It reports and repairs nothing.
novox/hq ADR 0145. Eight tests; the consecutive-failure logic proved by reverting
it once. Not yet registered or assigned.
75 endpoints across 50 modules, named from what each one is for rather than by a
rule: mail's seven protocol ports are smtp, imaps, submission and the rest; unifi's
nine are inform, stun, discovery, the two portal ports and syslog; minio's two are
s3 and console; the resolver's two are dns-udp and dns-tcp.
And 35 route contributions name the endpoint they serve instead of repeating its
port. A route and a listen both carried a port and nothing said they were the same
thing; now one of them does. gitea's path-level deny rule names neither, because it
is a rule about a name rather than an endpoint.
novox/hq ADR 0138. The words shipped a release ahead in mesh-controller #138 and
#139, and the control plane running today is built from that merge — checked before
this was written, because an unknown manifest key is refused and a catalogue using
one against an older control plane would stop resolving.
umami declared its port reachable from anywhere, reasoning that the collection
endpoint tracked browsers POST to must be public. That is true of the name and
not of the port: both its surfaces are served through the proxy by name, so the
port is how the proxy reaches it and nothing else (ADR 0045).
Measured, which is how this was found: with the port open to the internet, the
dashboard's login page was served over plain HTTP directly on the machine's port,
bypassing every rule the proxy applies by path. The route stays exactly as it was,
so the collection endpoint keeps working.
It declared udp/53 only. The daemon listens on tcp/53 as well, and a resolver is
asked over tcp whenever an answer will not fit in a datagram — so on every
converged machine that port is closed while the service reports itself healthy
and the manifest reads as though the resolver were fully declared.
The same fault as issue 136 in miniature: the declaration covers part of what the
service does, and the gap is silent because nothing compares the two.
The module said the service must be running and nothing about boot, so the
machine's own way back in was enabled only because something before the mesh
had enabled it. All four machines happen to be enabled today; none of them is
enabled because the mesh says so, and a machine adopted tomorrow would run ssh
until its first reboot.
Not `state: running` alone for the same reason the module exists: this is the
one daemon whose absence cannot be fixed remotely.
jail.local named ufw as the ban action. Two machines on this mesh have no ufw,
and fail2ban does not check: it starts, the jail reads the log, counts the
attempts, runs the ban command, gets 127 -- 'ufw: command not found' -- and
logs an error nobody reads. The service is active, the mesh reports the module
applied, and the machine is not protected. Proven by banning a documentation
address on such a machine today.
The replacement is this module's own dualchain action, already used by the
recidive jail on all four machines, so it is not a new dependency. It bans in
DOCKER-USER as well as INPUT, which ufw's action did not, and it bans all
ports, which ufw's action did.
The recidive jail bans whoever keeps coming back by reading fail2ban's own
log, and fail2ban checks every jail's log file while it configures itself --
before it has created that log. On a machine where the file is not there
already, no jail is found for recidive, configuration fails, and the whole
service refuses to start, taking the sshd jail with it. Two machines assigned
this module today came up failed for exactly that reason; the two where it
worked had a log from years of the service running.
Declared create-once: the mesh puts an empty file there when it is absent and
never touches it again, because what grows in it is fail2ban's, and the
logrotate file this module already ships is what keeps it small.
This also reverts the previous two commits' fail2ban.local. It declared a
logtarget that the package already sets to the same path on every machine
here -- pacman reports the config pristine -- so it fixed nothing and said
something untrue about why.
The file that says where fail2ban logs was not in restart-on, so a change to
it would sit on disk with the running service unaware of it -- the same shape
as any other jail file this module already restarts for.
The recidive jail reads /var/log/fail2ban.log and this module ships the
logrotate file for it, but nothing ever told fail2ban to write there. Where
the package default stands, fail2ban logs to the journal, the recidive jail
finds no log file, and the whole service refuses to start -- taking the sshd
jail with it. Two machines assigned this module today came up failed; the two
where it worked had /etc/fail2ban/fail2ban.conf edited by hand, which a
package upgrade would have undone.
Declared in fail2ban.local, because fail2ban.conf belongs to the package.
It asks what it missed on every start and the answer never arrived: the control plane replayed each
build it held as a module's event from a module called "control-plane", which does not exist, so its
own account refused the publish and the graph kept the gap. The control plane now states those under
the seat it holds (novox/hq ADR 0134, mesh-controller #129), so this consumes that too — one handler,
because what a build means for the graph is the same whether the build machine says it as it happens
or the mesh says what it already held.
It brought its schema up inside its runtime, on every start. That made a schema it could not reach a
crash loop rather than a stop, with the module graph keeping a gap and nothing saying so — which is
how a whole morning's builds went unrecorded. The mesh now prepares this module's state before it
starts this version and does not start it if that failed (novox/hq ADR 0135): the work moves to an
entrypoint the image names in MESH_PREPARE, beside the entrypoints it already names.
The reason it was at start — that a step blocking the apply would block the very apply bringing the
overlay up — stopped being true when a step's failure became its module's business rather than the
machine's (ADR 0136).
Its runtime announces that it has just started and may have missed builds — the event the control
plane follows to replay them — and its manifest did not declare it. A module's authority on the bus
is derived from what it declares, so the publish was refused and the runtime died on start, in a
loop, with the mesh's graph never catching up.
Every module built from a repository was rebuilt for a change to any of them: one merge in this
repository meant twenty-six builds, which is what exhausted a public registry's pull limit. The
forge lists the files a merge changed and the event carries them, from the watcher and from the
merge tool alike; a merge that changed more files than were asked for says so, and the mesh then
treats the whole repository as changed rather than guessing.
An old merge past the first page of the forge's listing surfaced as newer pull requests were
updated, and was announced as if it had just happened; the mesh then rebuilt everything built from
that repository, once per old merge. The moment the watching began is kept with the record, and
only a merge made since is announced.
A trigger that fires silently is indistinguishable from one that did not fire (novox/hq issue
131); the announcement is now one line an operator can read.
/user/repos lists what the token's user owns, which for the mesh's administrator is nothing — so
the forge module watched an empty list and never announced a merge. It reads the forge's whole
view through the search endpoint, every page.
The merge tool emitted at the instant it acted; a merge made in the forge's own
pages or over its API emitted nothing, and the mesh went on believing every module
current with its source (novox/hq 04-ISSUES/131). Merged pull requests are now
watched the way repositories are: what the forge holds, asked for on a tick,
announced once, with the merge commit and the clone URL a build needs. What has
been announced is kept beside the module's state, so a restart does not announce
the whole history again, and a first tick with no record announces nothing.
The mesh's broker certificate is the operator's, kept outside any module and
mounted read-only by whatever serves the bus; the controller declares that
access and now so does nats, the same way. A mount nothing declares is refused
at registration (ADR 0030), which is how this was found.
The module mounted a TLS directory nothing fills, so the server could not start
on a mesh that was not raised by the genesis template. The mesh already has a
broker certificate every machine pins by fingerprint and the controller trusts;
serving the new bus with it means no pin changes when a machine moves and there
is no second certificate to be wrong about. No ca_file: the directory has none,
and a pinning client checks the leaf and nothing else.
The recipe started FROM the upstream server's digest directly, and the build machine
refuses that: every base is declared under build.on and copied into the mesh's own
store before a build, so a build never reaches out to a registry the mesh does not
run (novox/hq ADR 0097). Found the first time the module was built on a real mesh.
Same digest, now declared as NATS_BASE and arriving as a build argument; the
Dockerfile says where it comes from and why the digest is the index's.
mesh-admin
jschoubben