Qualify peer artifact rollback under ENOSPC

[abhishek-anand]

Summary

• add an opt-in test failpoint after peer artifact disk reservation
• drive an isolated node filesystem below the required localization size using observed free space
• require a private 503 response with no partial artifact, snapshot, branch, replica, or staging state
• verify the source VM remains healthy and a clean retry completes cross-node restore
• detach the test loop device with the supported losetup --detach invocation

Validation

• cargo test --locked -p taritd --features test-failpoints (383 passed)
• strict clippy, formatting, shell syntax, and shellcheck
• exact c8i tree c7a12de7ed677968e151d6b5df2de015739c4d68
• Ubuntu 24.04 OCI E2E on Linux 6.6.155 and Linux 5.10.230
• real KVM, TAP networking, authenticated peer transfer, live fork, lazy restore, ENOSPC rollback, scale-to-zero, HTTP wake, durable egress, shared-volume recovery, replica repair, physical cleanup, and terminal teardown

[Copilot]

Pull request overview

This PR extends the peer artifact replication E2E gate to deterministically exercise (and verify) rollback behavior when node-B runs near ENOSPC during peer artifact localization, using an opt-in failpoint that pauses immediately after disk reservation.

Changes:

• Add a test-failpoints-gated taritd failpoint to pause after artifact localization disk reservation.
• Enhance the E2E script to mount node-B onto a loop-backed btrfs filesystem, exhaust free space below the required localization size, and assert a private 503 with no partial replica/snapshot/branch/staging state.
• Improve teardown by using the supported losetup --detach loop device cleanup path.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
orch/tests/e2e_peer_artifact_replication.sh	Adds near-ENOSPC fixture (loop+btrfs), drives storage pressure during localization pause, validates lossless rollback + clean retry, and updates cleanup behavior.
orch/crates/taritd/src/ops.rs	Introduces an opt-in localization pause failpoint immediately after disk reservation (gated behind test-failpoints).

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.