-
Notifications
You must be signed in to change notification settings - Fork 33.5k
Operator follow-up: finish HERMES_HOME/profile isolation smoke quirks #4671
Copy link
Copy link
Open
Labels
P3Low — cosmetic, nice to haveLow — cosmetic, nice to havearea/configConfig system, migrations, profilesConfig system, migrations, profilescomp/cliCLI entry point, hermes_cli/, setup wizardCLI entry point, hermes_cli/, setup wizardcomp/cronCron scheduler and job managementCron scheduler and job managementcomp/gatewayGateway runner, session dispatch, deliveryGateway runner, session dispatch, deliverytype/refactorCode restructuring, no behavior changeCode restructuring, no behavior change
Metadata
Metadata
Assignees
Labels
P3Low — cosmetic, nice to haveLow — cosmetic, nice to havearea/configConfig system, migrations, profilesConfig system, migrations, profilescomp/cliCLI entry point, hermes_cli/, setup wizardCLI entry point, hermes_cli/, setup wizardcomp/cronCron scheduler and job managementCron scheduler and job managementcomp/gatewayGateway runner, session dispatch, deliveryGateway runner, session dispatch, deliverytype/refactorCode restructuring, no behavior changeCode restructuring, no behavior change
Type
Fields
Give feedbackNo fields configured for issues without a type.
Summary
Backfill smoke still found a small set of non-blocking operator-facing inconsistencies around profile/HERMES_HOME isolation. This is not a release blocker, but it should be tracked as one bounded cleanup lane so operators can trust profile-scoped behavior.
Adjacent issues already exist for broader gateway/profile isolation problems (#4402, #4426, #4587). This follow-up is for the remaining smoke quirks below.
Scope
hermes gateway statusand global process visibility can still report/process data outside the activeHERMES_HOME, even when smoke is run against isolated homes.HOME-based rather thanHERMES_HOME-based, which makes profile-local expectations ambiguous.cron remove <missing-job>prints a failure-style message but exits 0, which is awkward for scripts/operators that rely on exit status.Why this matters
These do not currently block backfill smoke, but they weaken operator confidence in:
Desired outcome
Treat this as a bounded cleanup/polish pass:
cron removereturn a clearly actionable nonzero exit code when the target job does not exist (or intentionally document/standardize a different contract)Verification expectations
A fix here should include targeted coverage showing:
HERMES_HOMEscopeHERMES_HOMEsemanticsPriority
Non-blocking follow-up from backfill smoke. Worth cleaning up soon, but lower priority than operator-breaking isolation regressions or data-loss bugs.