Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

design: pending timeout for pods that never schedule, and image pull failures on the infra rail

Open
  • Is something wrong?
#1,609 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
5/5
Estimated time
Over a week
Newbie friendliness
15/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
go, kubernetes
Domain
backend, infrastructure

Research direction

Start in internal/executor/reconcile.go around lines 49-87, where Pending and the image-pull reasons are settled, then read pod_terminate.go:315 and stale_queued_reap.go:194, which defer on Pending. Read ADR 0051 and ADR 0053 for the intended infra rail and pending bound. This issue is a design proposal that depends on the infra settle path from the eviction issue, so agree on the design with maintainers before coding. Done means an unschedulable pod is failed after the configured timeout and image pull errors retry before settling.

Written by the indexing model from the issue text.

Description

area/executor enhancement priority/p3

Found in the 2026-10-09 performance and reliability re-audit (findings R-08 and R-09). Both reachable with default settings.

Pods that never schedule stay queued forever

internal/executor/reconcile.go:87 treats Pending as non-terminal unless a container waits on an unrecoverable reason, and an unscheduled pod has no container status. pod_terminate.go:315 and stale_queued_reap.go:194 treat Pending as a slow start and defer. activeDeadlineSeconds counts from status.startTime, which an unscheduled pod never gets, and nothing reads PodScheduled=False. A typo in node_selector, a missing toleration, a request larger than any node or a PVC that never binds leaves the task queued indefinitely, holding its pool slot and max_active_tasks. ADR 0053 asks for a bound here. Throwaway test on main 024196b9: an Unschedulable pod is still unsettled after 6 hours.

Image pull failures spend user retries

reconcile.go:49-50, 64-84 settle ErrImagePull and ImagePullBackOff as a task failure on the first observation. A registry 429 or 503 or a short network blip therefore fails the task and consumes retries (0 by default), although ADR 0051 puts bad images on the infra rail. Throwaway test: settled=true kind=failed on the first sweep.

Proposal

  • A setting such as executor.pending_timeout (default 10 to 15 minutes): a pod still Unschedulable past it is deleted and the task failed with a reason that names scheduling (infra re-place for capacity reasons if the operator prefers waiting).
  • Settle ErrImagePull and ImagePullBackOff as infra after a bounded number of back-offs, keep InvalidImageName terminal.
  • Both need the infra settle path in the reconciler that the eviction issue introduces.
Dominant language
Go
Stars
46
Forks
1
Avg merge
13h 15m
Merged PRs (30d)
642

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from dexadata/dexaflow

All issues in dexadata/dexaflow

Similar issues

More Go issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.