Fix get available memory to walk cgroup v2 ancestor chain

[preinlein]

What does this PR do?

Rewrites get_available_memory() to walk the cgroup v2 ancestor chain from the process-specific path upward, returning the tightest memory.max limit. The previous implementation only checked the cgroup root, which returns "max" when the container has its own cgroup namespace — causing lading to use u64::MAX instead of the actual container memory limit.

Motivation

In cgroup v2 environments where the container has its own cgroup namespace (common in Docker/Kubernetes), /sys/fs/cgroup/memory.max at the root reads "max" even though a real limit exists at the process-specific cgroup path. This caused lading to believe it had unlimited memory.

Related issues

N/A

Additional Notes

• Pins CI runners from ubuntu-latest to ubuntu-24.04 to speed up CI times.
• Introduces an injectable file reader (get_available_memory_with) for deterministic unit testing of cgroup parsing logic.
• Walks the full ancestor chain and takes the minimum limit, matching how the kernel enforces hierarchical memory limits.

[preinlein]

• (locally) Build & push lading container script #1827
• Adds SMP tests for idle & ascii #1826
• Fix get available memory to walk cgroup v2 ancestor chain #1824 👈 (View in Graphite)
• Add SMP CLI to preflight checks #1823
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[preinlein]

Note: This reduced the wait time for GH runners from ~15 min to seconds.

[preinlein]

I could see an argument for an info log in this function for what it resolved from where. Let me know if that's something you find desirable.

[preinlein]

Container build errors are fixed in #1827

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3d782ef017

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: cc851e7dbd

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[preinlein]

I need to make a fresh build and test this further. Codex kept finding cgroup edge cases so I want to at least test our current setup (lading as is and lading as a target).

[GeorgeHahn]

I think the code looks good, but I'm hesitant to take on the complexity of cgroupv1 support. What environment needs this? Can we switch it to cgroupv2 instead?

[preinlein]

I think the code looks good, but I'm hesitant to take on the complexity of cgroupv1 support. What environment needs this? Can we switch it to cgroupv2 instead?

Let me run a test. lading as target was misreporting memory but that shouldn't be a cgroupv1 issue. Claude just detected it as a possibility. I'm with you that if we don't need it, we shouldn't have it.

I'll drop it and run a test.

[preinlein]

I think the code looks good, but I'm hesitant to take on the complexity of cgroupv1 support. What environment needs this? Can we switch it to cgroupv2 instead?

Can confirm, it was just our cgroupv2 implementation was wrong.

Here are the logs with a fresh lading without any of the cgroupv1 stuff but with the changes to the cgroupv2: https://app.datadoghq.com/logs?query=job_id%3Af798fb29-9ba6-4bc7-9062-9248e739ff56%20%22amount%20of%20memory%22&agg_m=count&agg_m_source=base&agg_t=count&cols=host%2Cexperiment&messageDisplay=inline&refresh_mode=paused&storage=hot&stream_sort=time%2Cdesc&viz=stream&from_ts=1773919188864&to_ts=1773922788864&live=false

We can see that the memory reported is exactly what we expect.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b573bc91e4

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[preinlein]

Merge activity

• Mar 19, 1:25 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Mar 19, 1:27 PM UTC: Graphite rebased this pull request as part of a merge.
• Mar 19, 2:03 PM UTC: @preinlein merged this pull request with Graphite.