Skip to content

docs(podman): document macOS socket path mismatch and dynamic lookup - #1

Open
gmenher wants to merge 3 commits into
mainfrom
docs/macos-podman-socket-1690
Open

docs(podman): document macOS socket path mismatch and dynamic lookup#1
gmenher wants to merge 3 commits into
mainfrom
docs/macos-podman-socket-1690

Conversation

@gmenher

@gmenher gmenher commented Aug 4, 2026

Copy link
Copy Markdown
Owner

Summary

Documents the macOS Podman socket path mismatch that affects Homebrew-installed
Podman and provides the OPENSHELL_PODMAN_SOCKET workaround using
podman machine inspect.

Related Issue

Fixes NVIDIA#1690

Changes

  • docs/reference/sandbox-compute-drivers.mdx: Added "macOS Podman Socket Path"
    section documenting the issue and fix
  • .agents/skills/debug-openshell-cluster/SKILL.md: Added macOS socket mismatch
    diagnostic to Step 5 (Podman-Backed Gateways)

Testing

  • Verified default socket path does not exist on macOS Homebrew Podman
  • Verified podman machine inspect returns correct dynamic socket path
  • Verified resolved socket file exists and is accessible
  • mise run pre-commit passes
  • Unit tests added/updated
  • E2E tests added/updated (if applicable)

Checklist

  • Follows Conventional Commits
  • Commits are signed off (DCO)
  • Architecture docs updated (if applicable)

SDAChess and others added 3 commits August 4, 2026 15:38
Repeated Bazel test targets with --runs_per_test at 10, 20, 50, 100, and
200 runs. The tests failed intermittently when multiple instances ran
concurrently. This indicates timing, shared tracing state, socket readiness,
or parallel-safety issues that need focused follow-up before re-enabling the
tests.

A normal non-repeated bazel test //... run passes with these tests ignored.

Disabled tests:
- sandbox_forward_foreground_fails_when_ssh_exits_before_listener_opens
- sandbox_forward_background_terminates_owned_child_when_listener_never_opens
- podman_socket_probe_accepts_successful_ping_response
- podman_socket_probe_rejects_docker_ping_response
- docker_socket_probe_accepts_successful_ping_response
- docker_socket_probe_rejects_podman_ping_response
- docker_socket_detection_returns_the_responsive_candidate
- podman_socket_detection_returns_the_responsive_candidate
- driver_watch_events_are_roots_and_store_operations_have_parents
- reconcile_sweeps_are_roots_and_operations_have_parents
- gateway_listeners_bind_ipv6_wildcard_and_ipv4_callback_on_same_port
- watch_producer_releases_request_span_when_client_disconnects
- expected_conflicts_leave_the_span_unmarked
- store_spans_record_what_they_touched_as_attributes
- store_operations_export_spans_with_parents
- refresh_worker_ticks_are_roots_and_store_operations_have_parents

Signed-off-by: Simon Scatton <sscatton@nvidia.com>
* feat(build): add system CA root mode

Allow distro builds to use native trust stores for supervisor upstream TLS while keeping bundled Mozilla roots as the default. Avoid bundled root crates in system-ca-roots builds by using native-root TLS features and z3 0.20.

Signed-off-by: Adam Miller <admiller@redhat.com>

* fix(build): keep CA root feature in telemetry-off verification

The telemetry-off task uses --no-default-features which now disables
bundled-ca-roots in addition to telemetry, triggering the compile_error
guard. Re-enable bundled-ca-roots explicitly so the task verifies only
telemetry compilation.

Signed-off-by: Scott Burdine <sburdine@nvidia.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* fix(sdk): disable oauth2 default features to prevent webpki-roots leak

The bare `oauth2 = "5"` dependency re-enabled default features
(rustls-tls → reqwest/rustls-tls → webpki-roots), defeating the
system-ca-roots feature gate. Mirror the CLI fix: disable defaults
and enable only the `reqwest` feature.

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* refactor(build): simplify CA root selection to single feature toggle

Replace mutually exclusive bundled-ca-roots / system-ca-roots features
with a single bundled-ca-roots toggle. Disabling it implies system roots
via rustls-native-certs, which is now a regular (non-optional) dependency.
This fixes cargo --all-features and simplifies the distro build interface
from --no-default-features --features system-ca-roots to just
--no-default-features.

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* feat(build): add system-ca-roots convenience alias and fix verify task

Add a system-ca-roots feature alias on openshell-sandbox that includes
all other defaults (telemetry) except bundled-ca-roots, so distro
builds can use --no-default-features --features system-ca-roots without
manually re-adding unrelated defaults. Update the verify CI task to use
the alias and scope checks to the sandbox package. Fix task description
to use "build mode" terminology instead of implying a Cargo feature.

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* ci: fix system CA roots step name to use build mode terminology

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* refactor(sandbox): reorder features to place system-ca-roots alias near default

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

* fix(proxy): unwrap Result from build_upstream_client_config in tests

The function signature changed to return Result but the test call sites
were not updated, causing type mismatch compilation errors in CI.

Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Signed-off-by: politerealism <burdcat17@gmail.com>

---------

Signed-off-by: Adam Miller <admiller@redhat.com>
Signed-off-by: Scott Burdine <sburdine@nvidia.com>
Signed-off-by: politerealism <burdcat17@gmail.com>
Signed-off-by: Quinn Burdine <sburdine@redhat.com>
Co-authored-by: Adam Miller <admiller@redhat.com>
On macOS, Homebrew-installed Podman does not create the default socket
path that the Podman driver probes. Document the OPENSHELL_PODMAN_SOCKET
override and the podman machine inspect lookup in both the compute
drivers reference and the debug-openshell-cluster skill.

Fixes NVIDIA#1690

Signed-off-by: Gaizka Menendez Hernandez <gmenende@redhat.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

docs: document macOS Podman socket path mismatch and dynamic lookup

3 participants