vinzenz/hyperhive
vinzenz/hyperhive
Watch
0
0
Fork
You've already forked hyperhive
 0
Code Activity

give openbao's own start the container's long timeout

Browse source
This commit is contained in:
atlas 2026-11-01 15:03:28 +01:00
commit 8d31f0ad27
2 changed files with 51 additions and 30 deletions
Download patch file Download diff file Expand all files Collapse all files

18
nix/host-modules/swarm-bao.nix
View file

@ -52,6 +52,13 @@ let
# `nix/module-eval/grafana.nix` pins it against the shipped dashboard file. # `nix/module-eval/grafana.nix` pins it against the shipped dashboard file.
collectorServiceName = "bao"; 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;
# Upstream's own default, kept so its documentation matches. The raft data # Upstream's own default, kept so its documentation matches. The raft data
# lives INSIDE the container on `ephemeral = false`, the same way # lives INSIDE the container on `ephemeral = false`, the same way
# ./swarm-grafana.nix keeps its sqlite database — no sibling service binds # ./swarm-grafana.nix keeps its sqlite database — no sibling service binds
@ -4138,7 +4145,7 @@ in
# The container is `Type=notify`, so its start job waits for OpenBao, # 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 # 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. # storage. The 1min default kills it mid-open and every retry is cold.
timeoutStartSec = "5min"; timeoutStartSec = "${toString containerStartSec}s";
inherit privateNetwork; inherit privateNetwork;
@ -4395,8 +4402,13 @@ in
# attrset literal and does not parse. Module merging happens across # attrset literal and does not parse. Module merging happens across
# `config` blocks, not within a literal — so the seal's half joins # `config` blocks, not within a literal — so the seal's half joins
# with `optionalAttrs`. # with `optionalAttrs`.
systemd.services.openbao.serviceConfig = systemd.services.openbao.serviceConfig = {
lib.optionalAttrs haveServerTls { # Upstream sets no start limit, so the 90 s default cut off
# the 60-144 s raft open; `Restart=on-failure` upstream then
# restarts it cold.
TimeoutStartSec = openbaoStartSec;
}
// lib.optionalAttrs haveServerTls {
LoadCredential = [ "${serverKeyCredential}:${serverKeyPath}" ]; LoadCredential = [ "${serverKeyCredential}:${serverKeyPath}" ];
} }
# The PIN reaches openbao as an environment variable read from a # The PIN reaches openbao as an environment variable read from a

17
nix/module-eval/hive-tls.nix
View file

@ -853,10 +853,19 @@ let
) names; ) names;
} }
{ {
# The container is `Type=notify`: its start job lasts as long as # The container is `Type=notify`: its start job spans OpenBao's first
# OpenBao's first start, which a 1min limit cuts off mid-open. # start, so OpenBao's own limit is raised too and sits a minute below
name = "the store's container is given five minutes to start"; # the container's. `Restart=` comes from upstream, so it shows the
ok = lib.hasInfix "TimeoutStartSec=5min" allLocal.systemd.units."container@swarm-bao.service".text; # 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;
} }
{ {
name = "hive-priv is told the cache's address by the host's configuration"; name = "hive-priv is told the cache's address by the host's configuration";