Jobs & tasks

Two views of the same machinery. Understanding it explains most “why didn’t my change apply?” confusion.

The Dell job queue

Dell applies staged changes through jobs (IDs like JID_881783808368). Created implicitly when you:

• PATCH /Bios/Settings with ApplyTime OnReset (BIOS config job)
• Import an SCP (host config job)
• Push a firmware update

List: Managers/iDRAC.Embedded.1/Oem/Dell/Jobs (the idrac_jobs tool). The SAME JID is also visible as a standard Redfish task at /redfish/v1/TaskService/Tasks/{JID} — same state, two schemas.

Job lifecycle

Scheduled ──(trigger event, usually reboot)──> Running ──> Completed
                                                      └──> Failed (Message says why)

• Scheduled = waiting for its trigger. A BIOS config job sits here until the host reboots — it is NOT stuck.
• Running during POST for BIOS jobs (server takes minutes longer to boot; that’s the job executing).
• Failed carries the real validation error. Dell validates BIOS values at APPLY time — a successful idrac_bios_set + a Failed job means the value was rejected.

ApplyTime semantics

@Redfish.SettingsApplyTime on settings PATCHes:

• OnReset — apply at next reboot (what mcidrac uses; safe default)
• Immediate — only valid for things that don’t need POST (most BIOS attrs reject it)
• AtMaintenanceWindowStart — needs a configured maintenance window

DellAttributes (idrac_attributes_set) bypass all of this — they apply instantly with no job at all.

SCP task polling quirk

An SCP export task’s GET response morphs: while running it returns {"TaskState": "Running", "PercentComplete": N}; once complete, the SAME URI returns the exported configuration document with no TaskState at all. mcidrac’s poller treats a 200 with no task fields as the finished payload.

An SCP import with ShutdownType=NoReboot stays Scheduled until a manual reboot triggers it.

Cleaning up

A stuck job queue (rare; usually after interrupted firmware updates) can be cleared with racadm jobqueue delete --all over SSH, or the DellJobService.DeleteJobQueue OEM action. Doing this while a job is mid-flash can brick the component — only clear Scheduled/Failed jobs.