geomind_code/my_init
24
0
Fork 0
Code Issues 1 Pull requests Projects Releases Packages Wiki Activity Actions

feat/shim-and-cgroups #13

Closed
thabeta wants to merge 3 commits from feat/shim-and-cgroups into development
pull from: feat/shim-and-cgroups
merge into: geomind_code:development
Conversation 1 Commits 3 Files changed 31 +5685 -271
thabeta commented 2026-03-19 21:36:20 +00:00
Owner
Copy link
No description provided. 
Port from mos_init to my_init codebase.

Introduce zinit-shim, a long-lived per-service process monitor that
holds stdout/stderr pipes across zinit-server restarts. When the server
restarts, it reconnects to running shims via Unix sockets and resumes
log streaming.

Add cgroup v2 process tracking for ALL service types. The server creates
a per-service cgroup before spawning, assigns the child to it, and uses
cgroup.kill for reliable stop escalation. Solves double-fork escape and
provides atomic kill.

New crate: zinit_shim (crates/zinit_shim/)
New module: cgroup (crates/zinit_server/src/cgroup.rs)
New module: shim protocol (crates/zinit_sdk/src/shim.rs)

Config additions:
- service.shell (bool, default true): skip /bin/sh -c wrapper
- [resources] section: memory_max, cpu_weight (cgroup v2 limits)
- Port ~40 dependency graph edge-case tests from mos_init covering after,
  wants, conflicts, requires, mixed deps, cycles, removal, deep/wide
  graphs, diamond patterns, state transitions, and missing soft deps
- Add mkroot.sh and run.sh scripts for rootfs building and Docker/virtiofs
  launching
- Update Dockerfile: Alpine 3.23, zinit_shim binary, shim socket dir
- Rename zinit-shim binary to zinit_shim for naming consistency across
  all workspace binaries
feat: integrate youki libcgroups stats for detailed process observability
Some checks failed
Build and Test / build (pull_request) Failing after 1m21s
Details
Tests / test (pull_request) Failing after 1m24s
Details
780283e283
 
Vendor cgroup v2 stats-reading code from youki/libcgroups (Apache-2.0)
to provide comprehensive per-service observability: CPU usage/throttling/PSI,
memory current/peak/swap/cache/OOM events/PSI, block IO per-device bytes/ops/PSI,
and PID counts — all without pulling in youki's heavy dependency chain.

- Add cgroup_stats.rs: ~450 lines of pure parsing, 14 unit tests
- Add service.stats_detailed RPC method and GET /api/services/{name}/stats endpoint
- Add DetailedServiceStats SDK response types with PSI support
- Add stats_detailed() to both sync and async client libraries
- Refactor read_usage() to delegate to cgroup_stats::read_all_stats()
thabeta commented 2026-03-19 21:36:27 +00:00
Author
Owner
Copy link

done in f30b1c3d94

done in https://forge.ourworld.tf/geomind_code/my_init/commit/f30b1c3d94e0f359bd20a1d7717cc3aa8c6474dd
thabeta closed this pull request 2026-03-19 21:36:27 +00:00
Some checks failed
Build and Test / build (pull_request) Failing after 1m21s
Details
Tests / test (pull_request) Failing after 1m24s
Details

Pull request closed

Please reopen this pull request to perform a merge.
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
geomind_code/my_init!13
No description provided.