Image mounts linked to /manuals/dhi/troubleshoot.md, which moved to /manuals/dhi/how-to/troubleshoot.md in #25589 and broke the Hugo build.
Co-authored-by: Cursor <cursoragent@cursor.com>
Show the Engine 29.5 opt-in listener and matching client endpoint.
Explain group access and why the socket should use a user-profile
path.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
The distribution-provided docker-buildx package conflicts with Docker's official Buildx plugin on Debian and Ubuntu. Add it to the prerequisite package removal lists and commands while retaining the official docker-buildx-plugin package name.
Co-Authored-By: Claude <noreply@anthropic.com>
## Summary
Document the DSR option for Linux east-west overlay traffic and restore
the legacy overlay tutorial redirect.
Closes#9652
Generated by Codex
Co-authored-by: Codex <noreply@openai.com>
Resolver libraries can query multiple DNS servers differently, making results depend on the container network and runtime. Add a note distinguishing application-controlled lookups on the default bridge from Docker's ordered upstream lookups on custom networks.
Co-Authored-By: Codex <noreply@openai.com>
The read-only volume section incorrectly described the mount as a bind mount.
Refer to the mounted data as a volume and remove the host propagation claim.
Co-Authored-By: Codex <noreply@openai.com>
The seccomp guide did not explain where relative profile paths are resolved or which host must contain the file. Document that the Docker CLI resolves and reads the path before sending the profile contents to the daemon.
Co-Authored-By: Codex <noreply@openai.com>
The overlay networking docs omitted the supported DSR driver option. Document Linux east-west DSR usage and restore the legacy tutorial redirect.
Co-Authored-By: Codex <noreply@openai.com>
Document the image mount storage type: a new
content/manuals/engine/storage/image-mounts.md page following the
volumes/bind/tmpfs template, plus overview and cross-link updates in the
storage section.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds a "Use router-assigned IPv6 addresses" subsection to the macvlan
and ipvlan driver pages, covering how to receive IPv6 via SLAAC from a
router on the parent network instead of Docker IPAM.
When the network has no IPv6 subnet, Docker disables IPv6 on the
endpoint; setting the per-interface `net.ipv6.conf.IFNAME.disable_ipv6`
sysctl to `0` via the `endpoint.sysctls` driver option re-enables it so
the interface accepts router advertisements. The section links to the
`docker network connect` sysctl reference for the full mechanism. Also
adds `ipv6`/`slaac` keywords to both pages for discoverability.
Motivated by
[moby/moby#52815](https://github.com/moby/moby/issues/52815), where the
per-endpoint sysctl mechanism was hard to discover from the driver docs.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
Add a "Use router-assigned IPv6 addresses" subsection to the macvlan and
ipvlan driver pages explaining how to receive IPv6 addresses via SLAAC
instead of Docker IPAM. Setting the per-interface net.ipv6.conf.IFNAME.disable_ipv6
sysctl to 0 via the endpoint.sysctls driver option re-enables IPv6 on the
interface so it can accept router advertisements, and the section links to
the docker network connect sysctl reference for the full mechanism.
Also add ipv6 and slaac keywords to both pages to improve discoverability.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
## Description
- surface the cgroup v2 limitation earlier in the runtime metrics page
- clarify that the detailed metrics section currently documents the
cgroup v1 file layout
- point cgroup v2 readers to the container-specific controller files
they should inspect
## Related issues or tickets
- Addresses #24644
## Guideline alignment
- followed `docs/CONTRIBUTING.md`
- kept the change to one documentation file and one focused issue
## Validation
- not run (`docker buildx bake validate`) because this is a docs-only
clarification
## Reviews
- [x] Technical review
- [ ] Editorial review
- [ ] Product review
Fixes#23664.
The rootless docs compared rootless to userns-remap only in terms of
daemon privileges, never mentioning that the container-to-host UID
mapping also differs. Adds a short bullet list after the existing
comparison to spell it out and note the practical impact on bind-mount
file ownership.
The opening note framed the page only as background information without
explaining what the page actually documents. Clarify that it describes the
classic storage drivers, while pointing containerd image store users to the
relevant operational guidance.
Closes#25093
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The port requirements were listed twice: as a bullet list under
Requirements and as a more detailed table under Create an overlay
network. Keep the detailed table under Requirements, where the port
requirements belong, and drop the redundant copy.
Fixes#25073
Adds warning about containerd using a separate storage path from the
Docker data directory.
Changes:
- Added IMPORTANT callout in "Disk space usage" section explaining that
containerd uses separate storage
- Warns users with custom data directories that they need to configure
containerd separately
- Links to containerd configuration documentation for data directory
setup
This prevents users from unexpectedly filling their root partition when
they assumed their custom data directory configuration would apply to
containerd.
Fixes#24010
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
