fix(daemon): keep systemd gateway running after child OOM...

Recent Commits to openclaw:main

snowzlm · 2026-06-23 · via Recent Commits to openclaw:main

Original file line number	Diff line number	Diff line change
`@@ -86,6 +86,7 @@ RestartSec=5`
`86`	`86`	`TimeoutStopSec=30`
`87`	`87`	`TimeoutStartSec=30`
`88`	`88`	`SuccessExitStatus=0 143`
	`89`	`+OOMPolicy=continue`
`89`	`90`	`KillMode=control-group`
`90`	`91`
`91`	`92`	`[Install]`
`@@ -130,6 +131,11 @@ cat /proc/<child-pid>/oom_score_adj`
`130`	`131`	Expected value for covered children is `1000`. The Gateway process should keep
`131`	`132`	its normal score, usually `0`.
`132`	`133`
	`134`	+The recommended systemd unit also sets `OOMPolicy=continue`. This keeps the
	`135`	`+Gateway unit alive when a transient child process is selected by the OOM killer;`
	`136`	`+the child command/session can fail and report its error without systemd marking`
	`137`	`+the entire gateway service failed and restarting all channels.`
	`138`	`+`
`133`	`139`	`This does not replace normal memory tuning. If a VPS or container repeatedly`
`134`	`140`	`kills children, increase the memory limit, reduce concurrency, or add stronger`
`135`	`141`	resource controls such as systemd `MemoryMax=` or container-level memory limits.

Original file line number	Diff line number	Diff line change
`@@ -23,6 +23,7 @@ describe("buildSystemdUnit", () => {`
`23`	`23`	`expect(unit).toContain("TimeoutStopSec=30");`
`24`	`24`	`expect(unit).toContain("TimeoutStartSec=30");`
`25`	`25`	`expect(unit).toContain("SuccessExitStatus=0 143");`
	`26`	`+expect(unit).toContain("OOMPolicy=continue");`
`26`	`27`	`expect(unit).toContain("StartLimitBurst=5");`
`27`	`28`	`expect(unit).toContain("StartLimitIntervalSec=60");`
`28`	`29`	`expect(unit).toContain("RestartPreventExitStatus=78");`

Original file line number	Diff line number	Diff line change
`@@ -81,6 +81,10 @@ export function buildSystemdUnit({`
`81`	`81`	`"TimeoutStopSec=30",`
`82`	`82`	`"TimeoutStartSec=30",`
`83`	`83`	`"SuccessExitStatus=0 143",`
	`84`	`+// Transient child processes may be selected by the OOM killer before the`
	`85`	`+// gateway. Keep the service running when that happens; the child surface is`
	`86`	`+// already responsible for reporting the failed command/session.`
	`87`	`+"OOMPolicy=continue",`
`84`	`88`	`// Keep service children in the same lifecycle so restarts do not leave`
`85`	`89`	`// orphan ACP/runtime workers behind.`
`86`	`90`	`"KillMode=control-group",`

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。原文来自 — 版权归原作者所有。