Watch
0
0
Fork
You've already forked hyperhive
0

swarm-bao: agent certificates issued by a store-generated agent CA

An agent's store identity was signed in swarm-controller's memory by a CA
a controller-host unit generated on disk, and the listener never trusted
that CA. Agent leaves now come from the store itself: a `pki-agents` PKI
mount whose root openbao generates internally, so the agent CA's key
never exists outside the store.

- swarm-bao-agent-pki (new, store host, as the bao granter): enables and
  tunes the mount, generates the root once (guarded on an empty issuer
  list, no replace branch), upserts the `swarm-agent` role (client
  certificates named `hive-agent-*` only, 90 days), caches the CA at
  /var/lib/swarm-bao-tls/agent-ca.pem and composes the listener bundle.
- The listener's tls_client_ca_file is a new listener-client-ca.pem
  (client-ca.pem, then the agent CA). Host cert-auth roles still pin
  client-ca.pem, so an agent leaf satisfies no host role. swarm-bao-certs
  composes the same bundle before openbao starts.
- openbao reads tls_client_ca_file only at start, so when the bundle
  changed after openbao started, swarm-bao-agent-pki restarts
  openbao.service in the container; under `seal = "shamir"` it prints
  the step instead. Once swarm-bao-certs has a cached CA, later boots
  start openbao with it and do not restart.
- The controller policy gains exactly `update` on
  pki-agents/issue/swarm-agent. mint_and_verify now asks that role for
  the leaf (the store generates the key), writes the agent's cert-auth
  role pinning the issuing CA bao returned, and writes the agent's
  policy as render_agent alone: the hive-shared queue credential stanza
  is gone.
- deploy.bao.agentPkiRoleName (must start `swarm-`, asserted with the
  other pki role names); swarm-controller gets
  SWARM_CONTROLLER_AGENT_PKI_MOUNT/_ROLE from the deploy.bao options.

Deleted: swarm-controller-agent-ca and its options (agentCaFile,
agentCaKeyFile), env, LoadCredential entries and assertion;
agent_identity's Authority, rcgen signing and validity window; the
rcgen and time dependencies of swarm-controller (rcgen leaves the
workspace); policy::render_agent_with_queue and its tests. The CN-prefix
assertion policy.rs said was owed is not: agent and host roles pin
different CAs.

Migration is re-creating each agent after deploy; that overwrites the
stale role and policy.

Closes #4756
This commit is contained in:
atlas 2026-09-27 19:30:38 +02:00
commit 6170e74a31
16 changed files with 894 additions and 925 deletions

View file

@ -44,6 +44,8 @@ let
# The store's units live inside its container, so the gates below have to
# look there rather than at the host's service set.
baoUnits = machine: machine.containers.swarm-bao.config.systemd.services;
tlsDir = "/var/lib/swarm-bao-tls";
cases = [
{
# The store's seal is spread over six gates — the stanza, the
@ -214,6 +216,71 @@ let
name = "the store advertises a cluster address";
ok = lib.hasPrefix "https://" ((baoSettings baoPkcs11).cluster_addr or "");
}
{
# The listener trusts the agent CA through a file no cert-auth role
# names. At least one listener verifies clients, so an empty set cannot
# pass.
name = "every client-verifying listener reads the listener-only bundle";
ok =
let
verifying = lib.filter (l: l ? tls_client_ca_file) (
lib.attrValues (baoSettings baoPkcs11).listener
);
in
verifying != [ ]
&& lib.all (l: l.tls_client_ca_file == "${tlsDir}/listener-client-ca.pem") verifying;
}
{
# The other half of keeping agents out of host roles: those roles pin
# the store CA alone. The positive count is the control that the text
# searched is the one the roles are written in.
name = "host cert-auth roles pin client-ca.pem and never the listener bundle or the agent CA";
ok =
let
scripts = lib.concatStrings (
lib.mapAttrsToList (
n: u: lib.optionalString (lib.hasPrefix "swarm-bao-" n) (u.script or "")
) baoPkcs11.systemd.services
);
in
lib.hasInfix "certificate=@${tlsDir}/client-ca.pem" scripts
&& !(lib.hasInfix "certificate=@${tlsDir}/listener-client-ca.pem" scripts)
&& !(lib.hasInfix "certificate=@${tlsDir}/agent-ca.pem" scripts);
}
{
# Both writers compose the bundle through one function, which refuses
# a result that does not begin with the store CA and replaces the file
# only when its bytes change.
name = "both bundle writers use the one composer, which keeps the store CA first";
ok =
let
s = baoPkcs11.systemd.services;
composes =
u:
lib.hasInfix "compose_listener_bundle() {" u.script
&& lib.hasInfix "\ncompose_listener_bundle\n" u.script
&& lib.hasInfix ''cmp -s -n "$(stat -c %s ${tlsDir}/client-ca.pem)" ${tlsDir}/client-ca.pem "$tmp"'' u.script
&& lib.hasInfix ''cmp -s "$tmp" ${tlsDir}/listener-client-ca.pem'' u.script
&& lib.hasInfix "mktemp -p ${tlsDir} " u.script;
in
s ? swarm-bao-agent-pki && composes s.swarm-bao-certs && composes s.swarm-bao-agent-pki;
}
{
# openbao reads its client-CA file only at start, so picking up the
# agent CA is a restart: automatic where the store unseals itself,
# printed where a human has to. The shamir arm is the control.
name = "the agent PKI unit restarts openbao under pkcs11 and only prints the step under shamir";
ok =
let
restart = ''systemctl --machine="$machine" restart openbao.service'';
p = baoPkcs11.systemd.services.swarm-bao-agent-pki.script;
sh = baoShamir.systemd.services.swarm-bao-agent-pki.script;
in
lib.hasInfix restart p
&& lib.hasInfix ''if [ "$written_us" -le "$started_us" ]; then'' p
&& !(lib.hasInfix restart sh)
&& lib.hasInfix "# then unseal" sh;
}
];
in
runGroup "bao-basics" cases