[tzachi-runai]

Adds PageBroker benchmarks, k3s smoke tests, Qwen GPU checkpoint/restore tests, and autonomous endurance runs in one follow-up PR after #715. Supersedes #636 and #564.

• Benchmarks filesystem/POSIX and S3/Model Streamer transfers against one broker, with verified restores, resource measurements, and bounded cleanup.
• Runs isolated smoke and endurance Jobs with durable reports, process identity checks, and owned-prefix S3 cleanup. Preserves the current file-based broker credentials and artifact addressing.
• Exercises both Qwen models with repeated restores. Creates the default model-cache PVC when needed and downloads the selected model; shared NFS caches remain supported and reusable caches are retained.
• Keeps benchmark, Qwen helper, and smoke/endurance checks in CI. Integration coverage stays in test(pagebroker): separate integration coverage from the core stack #715; the core PRs are unchanged.

Validation: 84 Qwen/framework tests, 66 benchmark tests, and all 190 smoke/endurance tests passed (11 TLS tests ran separately with the integration dependencies). Repository make test and make build passed. make check stops at the existing Go/linter export-data incompatibility. GPU/k3s execution was not run because the configured cluster hostname could not resolve; sustained benchmark campaigns were not run.

Based on #715. Refs #267.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}