diff --git a/nix/host-modules/swarm-bao.nix b/nix/host-modules/swarm-bao.nix index bef4cf57b..64f17c383 100644 --- a/nix/host-modules/swarm-bao.nix +++ b/nix/host-modules/swarm-bao.nix @@ -52,16 +52,6 @@ let # `nix/module-eval/grafana.nix` pins it against the shipped dashboard file. collectorServiceName = "bao"; - # OpenBao's start job limit inside the container, and the container's own, - # which is the same plus a margin: the container is `Type=notify`, so its - # start job spans OpenBao's and must outlast it, or the container's limit - # kills a start the unit would still finish. - openbaoStartSec = 240; - containerStartSec = openbaoStartSec + 60; - - # systemd's default stop limit, which bounds the stop half of a restart. - openbaoStopSec = 90; - # Upstream's own default, kept so its documentation matches. The raft data # lives INSIDE the container on `ephemeral = false`, the same way # ./swarm-grafana.nix keeps its sqlite database — no sibling service binds @@ -3953,10 +3943,9 @@ in Restart = "on-failure"; RestartSec = 30; # An attempt may generate a 4096-bit RSA root and then wait on a - # blocking restart of openbao: its stop (`openbaoStopSec`) and its - # start (`openbaoStartSec`), plus 120s for the key generation and - # the calls around it. 90 + 240 + 120 = 450s. - TimeoutStartSec = "${toString (openbaoStopSec + openbaoStartSec + 120)}s"; + # blocking restart of openbao: its stop and its start, 90s each by + # systemd's defaults. + TimeoutStartSec = "5min"; }; script = '' set -euo pipefail @@ -4149,7 +4138,7 @@ in # The container is `Type=notify`, so its start job waits for OpenBao, # whose first start after a restart takes 60-144 s opening the raft # storage. The 1min default kills it mid-open and every retry is cold. - timeoutStartSec = "${toString containerStartSec}s"; + timeoutStartSec = "5min"; inherit privateNetwork; @@ -4406,36 +4395,31 @@ in # attrset literal and does not parse. Module merging happens across # `config` blocks, not within a literal — so the seal's half joins # with `optionalAttrs`. - systemd.services.openbao.serviceConfig = { - # Upstream sets no start limit, so the 90 s default cut off - # the 60-144 s raft open; `Restart=on-failure` upstream then - # restarts it cold. - TimeoutStartSec = openbaoStartSec; - } - // lib.optionalAttrs haveServerTls { - LoadCredential = [ "${serverKeyCredential}:${serverKeyPath}" ]; - } - # The PIN reaches openbao as an environment variable read from a - # 0400 file the provisioning unit wrote — never as a value in this - # expression, which would render it world-readable into the store. - # `TPM2_PKCS11_STORE` is required because the store is not at the - # library's default location. - // lib.optionalAttrs (baoDeploy.seal == "pkcs11") { - EnvironmentFile = pinEnvFile; - Environment = [ "TPM2_PKCS11_STORE=${tokenStoreDir}" ]; - # `DynamicUser` implies `ProtectSystem=strict`, which leaves the - # bind mount read-only to this unit however it is owned, and the - # pkcs11 library opens its sqlite store read-write. - ReadWritePaths = [ tokenStoreDir ]; - # Two groups, two different jobs: the store's sqlite file, and - # the TPM device the seal opens through it. `CapabilityBounding - # Set=` is empty upstream, so there is no `CAP_DAC_OVERRIDE` to - # fall back on — group membership is the only way in. - SupplementaryGroups = [ - tokenGroup - tpmGroup - ]; - }; + systemd.services.openbao.serviceConfig = + lib.optionalAttrs haveServerTls { + LoadCredential = [ "${serverKeyCredential}:${serverKeyPath}" ]; + } + # The PIN reaches openbao as an environment variable read from a + # 0400 file the provisioning unit wrote — never as a value in this + # expression, which would render it world-readable into the store. + # `TPM2_PKCS11_STORE` is required because the store is not at the + # library's default location. + // lib.optionalAttrs (baoDeploy.seal == "pkcs11") { + EnvironmentFile = pinEnvFile; + Environment = [ "TPM2_PKCS11_STORE=${tokenStoreDir}" ]; + # `DynamicUser` implies `ProtectSystem=strict`, which leaves the + # bind mount read-only to this unit however it is owned, and the + # pkcs11 library opens its sqlite store read-write. + ReadWritePaths = [ tokenStoreDir ]; + # Two groups, two different jobs: the store's sqlite file, and + # the TPM device the seal opens through it. `CapabilityBounding + # Set=` is empty upstream, so there is no `CAP_DAC_OVERRIDE` to + # fall back on — group membership is the only way in. + SupplementaryGroups = [ + tokenGroup + tpmGroup + ]; + }; # ⚠️ Upstream sets `restartIfChanged = false` on this unit: a restart # SEALS the store and disconnects every client. The container around diff --git a/nix/module-eval/bao-basics.nix b/nix/module-eval/bao-basics.nix index 623937b15..8aa2a2918 100644 --- a/nix/module-eval/bao-basics.nix +++ b/nix/module-eval/bao-basics.nix @@ -255,20 +255,6 @@ let && !(lib.hasInfix "certificate=@${tlsDir}/listener-client-ca.pem" scripts) && !(lib.hasInfix "certificate=@${tlsDir}/agent-ca.pem" scripts); } - { - # The unit's blocking `systemctl restart openbao.service` stops (90 s) - # then starts openbao (240 s), so its own limit must exceed both or a - # slow cold open kills the job mid-restart. Control: openbao's own limit - # is read from the same container, so a wrong path fails the 240 arm. - name = "the agent CA unit's start limit outlasts the restart of openbao it waits on"; - ok = - let - limit = lib.toInt ( - lib.removeSuffix "s" baoPkcs11.systemd.services.swarm-bao-agent-pki.serviceConfig.TimeoutStartSec - ); - in - (baoUnits baoPkcs11).openbao.serviceConfig.TimeoutStartSec == 240 && limit >= 90 + 240 + 60; - } { # Both writers compose the bundle through one function, which refuses # a result that does not begin with the store CA and replaces the file diff --git a/nix/module-eval/hive-tls.nix b/nix/module-eval/hive-tls.nix index 92a95f61a..ca8575610 100644 --- a/nix/module-eval/hive-tls.nix +++ b/nix/module-eval/hive-tls.nix @@ -853,19 +853,10 @@ let ) names; } { - # The container is `Type=notify`: its start job spans OpenBao's first - # start, so OpenBao's own limit is raised too and sits a minute below - # the container's. `Restart=` comes from upstream, so it shows the - # lookup reaches the real unit. - name = "openbao's start limit is raised and stays below its container's"; - ok = - let - inner = allLocal.containers.swarm-bao.config.systemd.services.openbao.serviceConfig; - unit = allLocal.systemd.units."container@swarm-bao.service".text; - in - inner.Restart == "on-failure" - && inner.TimeoutStartSec == 240 - && lib.hasInfix "TimeoutStartSec=300s" unit; + # The container is `Type=notify`: its start job lasts as long as + # OpenBao's first start, which a 1min limit cuts off mid-open. + name = "the store's container is given five minutes to start"; + ok = lib.hasInfix "TimeoutStartSec=5min" allLocal.systemd.units."container@swarm-bao.service".text; } { name = "hive-priv is told the cache's address by the host's configuration";