From software problem to a clear next step.
Choose a topic and what you need to do next. Add a search phrase to find a specific problem.
Browse 20 topics
Explore the playbooks
1500 guides · Page 34 of 60
-
Triage job priority inversion from ready time to claim
Locate a priority inversion by comparing job readiness, declared priority, queue membership, worker eligibility, claim order, and actual start time.
-
Prioritize job priority inversion by blocked urgency
Rank priority dispatch work by the urgent action delayed, queue reach, duration of inversion, and confidence that policy rather than capacity caused the order.
-
Investigate priority dispatch with matched ready times
Build a reproducible priority brief using matched jobs, queue placement, worker eligibility, and trace evidence from readiness through claim and start.
-
Verify priority ordering across queues and worker classes
Check urgent and lower-priority pairs in quiet, busy, tied, and late-arrival conditions, confirming declared precedence survives queue and worker boundaries.
-
Document the job priority and queue boundary contract
Preserve lessons about priority propagation, queue-local ordering, worker eligibility, tie handling, and dispatch fixtures for future queue changes.
-
Triage low-priority job starvation by wait age and service
Determine whether favored arrivals starve low-priority work by measuring ready age, claim opportunities, worker capacity, and service share across a controlled interval.
-
Prioritize starvation by maximum wait and service consequence
Rank low-priority starvation work by oldest wait, blocked maintenance or customer outcome, affected class, and evidence that eligible service was repeatedly bypassed.
-
Investigate starvation with a controlled arrival stream
Reproduce low-priority starvation using fixed worker capacity, a sustained favored stream, and traces that expose eligibility, selection, claim, and service share.
-
Verify low-priority service under a sustained favored stream
Check quiet, mixed, and sustained favored workloads to confirm an eligible low-priority job eventually claims without obscuring urgent service behavior.
-
Set a recurrence guard for low-priority starvation
Document maximum wait expectations, eligible service opportunities, workload fixtures, and the queue changes that require repeating starvation checks.
-
Triage queue fairness drift with per-class service records
Locate fairness drift by comparing ready, claim, completion, wait age, and service share for each work class across the same queue interval.
-
Prioritize queue fairness drift by class impact
Rank fairness drift by the work class losing service, oldest wait, blocked outcomes, persistence across intervals, and evidence of an avoidable allocation change.
-
Investigate fairness drift with equal and skewed arrivals
Build a queue fairness brief by comparing equal-class and skewed-class arrivals, worker allocation, selector choices, and service share over matched intervals.
-
Verify queue fairness across equal and skewed workloads
Check per-class claim and completion behavior under equal, skewed, and changing arrivals, confirming oldest waits and service shares remain within the stated policy.
-
Record a per-class queue fairness contract
Preserve class-specific service expectations, oldest-wait signals, equal and skewed fixtures, and review triggers for queue or worker allocation changes.
-
Triage worker concurrency thrashing with resource traces
Find whether more worker concurrency reduces useful throughput by comparing fixed workloads, completion time, CPU, memory, database connections, locks, and queue age.
-
Prioritize concurrency thrashing by resource and queue risk
Rank worker concurrency work by lost useful throughput, queue growth, database or memory pressure, error amplification, and the cost of a bounded tuning experiment.
-
Investigate worker concurrency with a bounded sweep
Build a concurrency brief by sweeping controlled worker levels and correlating useful throughput, queue age, resource saturation, lock waits, and retries.
-
Verify worker concurrency at the selected resource ceiling
Check lower, selected, and higher worker concurrency against fixed jobs, resource headroom, queue age, retries, and terminal correctness before accepting a tuning change.
-
Document a worker concurrency ceiling by job class
Preserve concurrency limits, workload fixtures, resource guards, and the measurements that distinguish contention from downstream capacity for future worker changes.
-
Triage a missing heartbeat against long-job phase progress
Determine whether a long-running job is dead or merely quiet by comparing heartbeat timestamps, phase progress, monitor age, worker health, and terminal events.
-
Prioritize long-job heartbeat gaps by intervention risk
Rank heartbeat work by duplicate intervention, abandoned resource ownership, blocked dependents, and confidence that quiet work is actually still running.
-
Investigate long-job heartbeats during a slow phase
Reproduce heartbeat silence in a controlled long phase and trace sender, worker progress, monitor observation, lease, and terminal-state behavior.
-
Verify heartbeats for short, long, and blocked job phases
Check liveness during quick work, a slow phase, and a dependency wait, confirming heartbeat age, progress wording, worker ownership, and terminal decisions stay distinct.
-
Define phase-aware heartbeat expectations for long jobs
Preserve heartbeat cadence, phase progress, monitor interpretation, ownership evidence, and fixtures for long-running work that can legitimately remain quiet.