weight_lowering_age_seconds
For thefair workload scheduler: once the query has been running (wall-clock) for this many seconds, its weight is lowered by weight_lowering_factor. 0 (or any negative value) disables the age threshold.
weight_lowering_cpu_seconds
For thefair workload scheduler: once the query has attained this many CPU-seconds, its weight is lowered by weight_lowering_factor. Applies to CPU resources. 0 (or any negative value) disables the CPU threshold. Attained CPU is the granted scheduler service, charged when a request is granted rather than as CPU is spent, so it leads actual consumption by at most one quantum (cpu_slot_quantum_ns) per active slot. This is only meaningful with CPU slot preemption (cpu_slot_preemption = 1, the default); without preemption CPU slots carry a fixed per-slot cost rather than real CPU time, so this threshold — like the other fair CPU settings — no longer reflects actual CPU consumption.
weight_lowering_factor
For thefair workload scheduler: once the query crosses any of the weight_lowering_* thresholds below, its effective weight is multiplied by this factor once (values in (0, 1) lower the weight, biasing scheduling toward shorter/newer queries). The thresholds do not combine — the first one to trip applies the full lowering. 1.0 disables lowering. The value is clamped to the range [0, 1], so the factor can only ever lower a query’s weight, never raise it.
weight_lowering_io_bytes
For thefair workload scheduler: once the query has attained this many bytes of IO, its weight is lowered by weight_lowering_factor. Applies to IO resources. 0 (or any negative value) disables the IO threshold.