Harden CI deploy timeouts for heavier post-upgrade images - #38
Closed
klagrida wants to merge 2 commits into
Closed
Conversation
E2E Tests (Linux) started failing on main after #34 bumped the Spark images (3.5.0 -> 4.0.1). The larger images increase pull contention on the constrained minikube CI node, pushing the (unpinned, ~4GB) jupyter/pyspark-notebook:latest image past Terraform's default 10m rollout wait: Error: Waiting for rollout to finish: 1 replicas wanted; 0 replicas Ready with kubernetes_deployment_v1.jupyter The image was at 9m40s and still pulling when the wait expired, while Prometheus/Grafana each took ~5.5m on the same node, confirming it's slow image-pull, not a real deploy error. Add a 20m create/update timeout and bump progress_deadline_seconds to 1200 on the Jupyter deployment. Durable fix (pinning a smaller image) tracked in #36. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The #34 version bumps (Spark 3.5->4.0.1, Airflow 3.0.2->3.1.7) increased total image bytes pulled onto the single minikube CI node. With more pull contention, whichever deployment loses the image-pull lottery exceeds its timeout - the failure moved between runs (jupyter, airflow, prometheus, grafana) rather than being one broken component: - platform-tests run: airflow still pulling when the 15m step timeout hit - ci run: airflow deployed fine (6m17s) but prometheus/grafana hit the provider's default 10m rollout wait Give the heavier deploys consistent headroom: - prometheus + grafana: timeouts.create/update=20m, progress_deadline=1200 (matching the jupyter fix in the previous commit) - airflow helm_release: timeout 900 -> 1500 (25m) - Deploy step timeout-minutes: platform-tests 15 -> 30, ci 20 -> 30 Durable fix (smaller/pinned images, pre-pull) tracked in #36. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
klagrida
force-pushed
the
fix/jupyter-rollout-timeout
branch
from
June 6, 2026 01:13
010187e to
09ed132
Compare
klagrida
changed the title
Contributor
Author
|
Obsolete after #39 reverted the version-drift change. With the lighter baseline images restored, CI deploys complete in ~11–13min and no longer need the extended timeouts this PR added. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
After #34 (Spark 3.5→4.0.1, Airflow 3.0.2→3.1.7),
E2E Tests (Linux)and the Cluster Integration Test started failing onmain. Root cause is image-pull saturation, not a broken component: the upgrade increased the total image bytes pulled onto the single minikube CI node, so whichever deployment loses the pull lottery exceeds its timeout. The failure moved between runs:E2E was green on every commit before #34 and fails reproducibly after, so this is a real (timing-dependent) regression introduced by the heavier images.
Change — give the heavy deploys consistent headroom
kubernetes_deployment_v1.jupytertimeouts.create/update=20m,progress_deadline_seconds=1200kubernetes_deployment_v1.prometheuskubernetes_deployment_v1.grafanahelm_release.airflowtimeouttimeout-minutes(platform-tests)timeout-minutes(ci)Also adds
.claude/to.gitignore(harness state).Durable fix
These timeouts are a band-aid for slow image pulls. The real fix — pinning smaller/maintained images (esp. the ~4 GB
jupyter/pyspark-notebook:latest) and/or pre-pulling — is tracked in #36.Verification
terraform fmt/terraform validate🤖 Generated with Claude Code