[codex] Record exec-server lifecycle metrics

[richardopenai]

Summary

• Record bounded connection, request, and process lifecycle metrics.
• Serialize active-gauge updates so concurrent starts and finishes cannot publish stale values.
• Track process completion exactly once across normal exit, explicit termination, and session shutdown.
• Cover the recorder in memory and verify real OTLP/HTTP export through a codex exec-server subprocess.

Stack

Review and land this stack in order:

1. [codex] Trace exec-server JSON-RPC requests #27466 — trace exec-server JSON-RPC requests
2. [codex] Record exec-server lifecycle metrics #27467 — record bounded connection, request, and process lifecycle metrics (this PR)
3. [codex] Observe remote exec-server lifecycle #27470 — observe remote registration and Noise rendezvous lifecycle

Validation

• just test -p codex-exec-server --lib (147 passed)
• just test -p codex-cli --test exec_server (3 passed)
• just argument-comment-lint
• just bazel-lock-check
• just fix -p codex-exec-server -p codex-cli
• just fmt

[chatgpt-codex-connector]

Review source: Codex Cloud Agents (CCA)

Codex Cloud Agents (CCA) couldn't complete this review. The original Codex Review is unaffected.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d6b943d92c

ℹ️ About Codex in GitHub

Your team has set up Codex to 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 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Serialize active connection gauge updates

When two connections of the same transport start or finish concurrently, the atomic counter update and the gauge recording can be reordered: e.g. one disconnect computes active = 1, another computes and records 0, then the first records stale 1, leaving exec_server_connections_active overreported until another event. Please make the active gauge observe the atomic value at collection time or protect the update+record sequence so the emitted value cannot go backwards after a newer count was recorded.

Useful? React with 👍 / 👎.