[barnabasbusa]

The focil-devnet-1 lodestar image and dora master-latest were both rebuilt on Oct 12 with EIP-8015's container changes (deposit and eth1data fields removed), which devnet-1 doesn't run. Lodestar can't read its own archived heze state ("Cannot extract weak subjectivity summary from archived state bytes") so all six lodestar nodes crash-looped after watchtower pulled it and finality stalled at epoch 1360, and dora, redeployed by the hostname switchover, fails to decode every heze block. This pins lodestar to focil-devnet-1-1134cea and dora to master-1b62438, the last builds before those merges.

https://claude.ai/code/session_01B7h25qcCifoGJpJGMDge43

[redpandabot]

Pins the devnet-1 lodestar image from the moving focil-devnet-1 branch tag to focil-devnet-1-1134cea. I verified the tag convention against another ethpandaops devnet's images.yaml and confirmed via the upstream ChainSafe/lodestar repo that 1134cea was the focil-devnet-1 tip immediately before the Oct 12 merge that pulled in EIP-8015, so the pin is well-founded and correctly scoped. No real problems found.

---

_{Reviewed 1 changed file(s) @ 381c3460 — no blocking issues found.
"If you're not embarrassed by the first version, you launched too late." — Reid Hoffman}

[redpandabot]

Summary

Pins the devnet-1 lodestar image in ansible images.yaml and the dora image in the kubernetes dora values.yaml to pre-EIP-8015 builds. The lodestar pin is in the canonical source, but the dora pin is applied to an ansible-generated file while images.yaml still holds the broken master-latest, so the dora fix will not survive the next playbook run.

Issues

• 🔴 kubernetes/devnet-1/dora/values.yaml:7 — dora pin edited in an ansible-generated file; source images.yaml still says master-latest — see the thread on that line

---

_{Reviewed @ e49f4bf1
"If you're not embarrassed by the first version, you launched too late." — Reid Hoffman}

[redpandabot]

🔴 dora pin edited in an ansible-generated file; source images.yaml still says master-latest

This file is generated by the generate_kubernetes_config role (it carries the "Ansible managed. Don't change this file manually." header, and the collection's templates/dora.yaml.j2 builds the tag from default_tooling_images.dora), so the pinned tag is overwritten whenever ansible-playbook playbook.yaml runs its gen_kubernetes_config role. ansible/inventories/devnet-1/group_vars/all/images.yaml line 44 still reads dora: ethpandaops/dora:master-latest — the same source the PR correctly updated for lodestar — so the dora outage returns on the next playbook run unless that line is pinned too.