# Release 3.5.5 (/release-notes/v3-5-5)



**Released September 28, 2026.** The published server image index now labels its `amd64` child correctly, so an x86 host or a GKE node can pull it; a rebuilt node's boot identity claim retries instead of silently falling back to the wrong replica id; and kmq adds commands for administering one rebuilt server. The operator and the Helm chart are unchanged from [3.5.4](/release-notes/v3-5-4).

## Artifacts [#artifacts]



| Artifact                 | Version                                                                                            |
| ------------------------ | -------------------------------------------------------------------------------------------------- |
| kmq                      | `v3.5.5`                                                                                           |
| Server image             | `europe-docker.pkg.dev/kubemq/images/kubemq-next:v3.5.5`, for `linux/amd64` and `linux/arm64`      |
| Server image (FIPS)      | `europe-docker.pkg.dev/kubemq/images/kubemq-next-fips:v3.5.5`, for `linux/amd64` and `linux/arm64` |
| Operator image           | `europe-docker.pkg.dev/kubemq/images/kubemq-operator-next:v3.4.0` (unchanged)                      |
| Helm chart `kubemq-next` | 3.4.0 (unchanged; it installs server `v3.5.0`)                                                     |

## Before you upgrade [#before-you-upgrade]

* **Nothing new to prepare.** The operator and the Helm chart are the same as in 3.5.4.

## Changes [#changes]

* **The published multi-arch server image index no longer mislabels its `amd64` child as the `arm64` variant.** An x86 host or a GKE node could not pull `kubemq-next` or the FIPS image at all. Every child image is now added to the index with its architecture and operating system set explicitly, and a check refuses to publish an index shaped like the broken one again.
* **A rebuilt node's boot identity claim retries instead of giving up after 10 seconds.** It used to send one request before the node had route interest to its peers, then silently fall back to the configured replica id, so `kmq cluster replace` never actually took effect on Kubernetes. The claim now retries for 90 seconds with backoff, never falls back while a peer is reachable, and logs which one happened.
* **kmq adds commands for administering a rebuilt server.** Membership commands can sign in as the dashboard administrator for one operation (`--admin-username`, `--admin-password-stdin`); `kmq auth login --server-index N --bootstrap` recovers one rebuilt server; `kmq deploy verify --accept-rebuilt-server N` records that a server's volume identity was intentionally replaced. `kmq deploy prepare` now refuses an image with no entry for a node's own architecture instead of failing later, and `kmq deploy diagnose` classifies image-pull failures.

## Known issues [#known-issues]

* No default connector image is published. Set `spec.image` on every `KubemqConnector`.