REN-1017 / REN-1171: re-land of homelab#1183, reverted in #1188, with the fixed controller.
Stacked on homelab#1189 (device 30): both PRs append to the end of the register. Merge #1189 first; after that this PR shows only its own commit 8c377a2c.
What changed
Controller:main-1d03144 → main-a5b299d@sha256:ab1309f5…f4ba60 (mqtt-account-controller#8, REN-1171: dynsec commands at QoS 0, so a self-kick does not resend modifyClient on reconnect). I checked the digest against the registry.
Register: H&T 35–38 with wake: shellies/<hostname>/online. This is identical to #1183, which Argus checked on REN-1146: the controller creates the wake read role itself.
Impact
The controller pod restarts once. One expected broker kick when the wake role is added. HA and the broker do not restart, so it is not an outage merge.
Verified
kustomize build OK. After the merge, Atlas checks what Hephaestus specified:
Pod Ready and /healthz 200 for at least 3 intervals.
At most oneadministrative action right after adding wake role.
No listClients: publish: timeout after the first loop.
If it loops again: revert to main-1d03144 and open a card for Hephaestus.
REN-1017 / REN-1171: re-land of homelab#1183, reverted in #1188, with the fixed controller.
**Stacked on homelab#1189 (device 30):** both PRs append to the end of the register. Merge #1189 first; after that this PR shows only its own commit `8c377a2c`.
## What changed
- **Controller:** `main-1d03144` → `main-a5b299d@sha256:ab1309f5…f4ba60` (mqtt-account-controller#8, REN-1171: dynsec commands at QoS 0, so a self-kick does not resend `modifyClient` on reconnect). I checked the digest against the registry.
- **Register:** H&T 35–38 with `wake: shellies/<hostname>/online`. This is identical to #1183, which Argus checked on REN-1146: the controller creates the wake read role itself.
## Impact
The controller pod restarts once. One expected broker kick when the wake role is added. HA and the broker do not restart, so it is not an outage merge.
## Verified
`kustomize build` OK. After the merge, Atlas checks what Hephaestus specified:
- Pod Ready and `/healthz` 200 for at least 3 intervals.
- At most **one** `administrative action` right after `adding wake role`.
- No `listClients: publish: timeout` after the first loop.
If it loops again: revert to `main-1d03144` and open a card for Hephaestus.
## Revert
Revert the commit. It is the same shape as #1188.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Approved at 8c377a2cfd. Premise: REN-1171 cause (paho replays unacked QoS1 modifyClient on reconnect) fixed in a5b299d = 982231a + f237378 (QoS 0); includes REN-1115 wake-no-rotate-on-unreachable. Register lines identical to #1183 (099aea38). kustomize build home-assistant OK. CI green. Only the controller pod restarts (Recreate, 1 replica). Revert = revert commit.
Approved at 8c377a2cfd8bfe4d0c7f796a048e5ea9ed5854de. Premise: REN-1171 cause (paho replays unacked QoS1 modifyClient on reconnect) fixed in a5b299d = 982231a + f237378 (QoS 0); includes REN-1115 wake-no-rotate-on-unreachable. Register lines identical to #1183 (099aea38). kustomize build home-assistant OK. CI green. Only the controller pod restarts (Recreate, 1 replica). Revert = revert commit.
Ghost
merged commit d9bfaf0a65 into master2026-10-07 17:00:06 +00:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
REN-1017 / REN-1171: re-land of homelab#1183, reverted in #1188, with the fixed controller.
Stacked on homelab#1189 (device 30): both PRs append to the end of the register. Merge #1189 first; after that this PR shows only its own commit
8c377a2c.What changed
main-1d03144→main-a5b299d@sha256:ab1309f5…f4ba60(mqtt-account-controller#8, REN-1171: dynsec commands at QoS 0, so a self-kick does not resendmodifyClienton reconnect). I checked the digest against the registry.wake: shellies/<hostname>/online. This is identical to #1183, which Argus checked on REN-1146: the controller creates the wake read role itself.Impact
The controller pod restarts once. One expected broker kick when the wake role is added. HA and the broker do not restart, so it is not an outage merge.
Verified
kustomize buildOK. After the merge, Atlas checks what Hephaestus specified:/healthz200 for at least 3 intervals.administrative actionright afteradding wake role.listClients: publish: timeoutafter the first loop.If it loops again: revert to
main-1d03144and open a card for Hephaestus.Revert
Revert the commit. It is the same shape as #1188.
🤖 Generated with Claude Code
Approved at
8c377a2cfd. Premise: REN-1171 cause (paho replays unacked QoS1 modifyClient on reconnect) fixed in a5b299d = 982231a + f237378 (QoS 0); includes REN-1115 wake-no-rotate-on-unreachable. Register lines identical to #1183 (099aea38). kustomize build home-assistant OK. CI green. Only the controller pod restarts (Recreate, 1 replica). Revert = revert commit.