[Bug] Default Helm install fails: minio/minio was removed from Docker Hub

[rhabichl]

TL;DR

The minio/minio repository is gone from Docker Hub.
Every fresh helm install of the Milvus chart with the default
minio.enabled: true now lands in ErrImagePull / ImagePullBackOff.
Existing clusters keep running only while the image stays in cache,
any node replacement, scale-out, or eviction takes the pod down.

Background

• The upstream minio/minio repository was archived in late April 2026. community edition is source-only from that point on.
• Docker Hub distribution has since been withdrawn entirely.
• quay.io/minio/minio still serves the old tags, but is not maintained.

So repointing at quay.io restores installs but leaves the chart shipping an outdated image with potential known unpatched CVEs.

Workaround in use

Patched the Deployment image to pgsty/silo (Silo, the
community-maintained MinIO server fork).
Existing data volumes were reused as-is with no migration.

One change was required: Silo renames the executable, so the command
must call silo instead of minio.
The S3 API, MINIO_* env-vars, minio_* metrics, /minio/* routes and the on-disk format are
unchanged, so Milvus itself needed no configuration change.

Proposed options

• Switch the bundled subchart to Silo (pgsty/silo + pgsty/mc). Keeps
  the wire protocol and on-disk format, so it is an image swap rather than a
  migration, and rollback is a tag revert. Requires the binary-name change in
  the subchart templates and a healthcheck review (mcli vs mc).
• Switch to a different S3 implementation like RustFS
• Demote bundled object storage entirely and make externalS3 the
  documented default,.

Questions for maintainers

• Which of the above suits you? I have a PR in progress for option 1 and I am
  happy to divert if you'd prefer a different direction.