Convert LoadBalance.actor.h to standard coroutines

[tclinkenbeard-oai]

Summary

• Convert load-balancing, replica-comparison, and TSS-comparison actors to standard C++ coroutines.
• Consolidate the migrated implementation into fdbrpc/LoadBalance.h and update direct includes so the header no longer requires actor compilation.
• Preserve request ownership, cancellation, inert Never() handling, race ordering, replica consistency checks, and Void-reply compatibility.

Testing

• fdbrpc_test -f /fdbrpc/waitValueOrSignal/ — 3 tests passed.
• fdbclient_test -f /StorageServerInterface/TSSCompare/ — 1 test passed.
• Built fdbserver successfully.
• tests/fast/CycleTest.toml with replica consistency checking enabled — 2 simulation tests passed.
• clang-format and git diff --check passed.

[tclinkenbeard-oai]

Generated by Codex.

What is it trying to do?

Convert FoundationDB’s load-balancing, replica-comparison, and testing-storage-server comparison actors to standard C++ coroutines. Consolidate the implementation into LoadBalance.h, update consumers and documentation, and add regression tests for completed-request ownership.

Is it correct?

The migration preserves the important behavioral contracts: request retries, cancellation, at-most-once delivery handling, failure-monitor integration, speculative second requests, queue-model accounting, replica comparison, and deterministic race ordering.

Both load-balancer implementations move their owning request and server-team parameters into coroutine-body locals, ensuring those resources are released before completion callbacks run. The new regression tests exercise real WaitMetricsRequest objects and verify that completed futures no longer retain server-team references or reply promises.

No serialized fields, persisted formats, or protocol compatibility boundaries are changed.

Current-head clang-format and clang-tidy have passed. The Windows check failed while downloading Boost with HTTP 503, before FoundationDB was configured or built; the same dependency failure is visible on unrelated changes and upstream main, so it does not indicate a defect in this PR.

This review was source-only; I did not run builds or tests.

Are there bugs?

I did not find any remaining correctness bugs.

Are there omissions?

The new regression tests cover successful, single-server request completion but do not directly exercise cancellation during an outstanding request, retryable errors, at-most-once failures, or speculative second requests. These are residual coverage gaps rather than demonstrated defects.

Two references to the deleted actor header also remain in design/AI-generated/subsystem_02_rpc_transport.md; the primary hand-written load-balancing design document has already been corrected.

Are there better ways of doing things?

race(firstRequestData.response, secondDelay) materializes an ErrorOr<Reply> inside its result variant even though this call site only uses the winning index and rereads the response separately. An index-only completion helper could avoid that reply copy on the read path if profiling demonstrates that it matters. The previous implementation also allocated a helper coroutine, so the current code does not establish an additional allocation by itself.

Should this CL be LGTMd?

I would wait for the current-head FoundationDB builders and cluster tests to complete, then LGTM if they pass. The outstanding Windows failure is an independently verified external Boost-download problem, not a source-level correctness blocker.

[foundationdb-ci]

Result of foundationdb-pr-clang-arm on Linux RHEL 9

• Commit ID: 241ad54
• Duration 0:45:16
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr-clang-ide on Linux RHEL 9

• Commit ID: 241ad54
• Duration 0:53:26
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr-macos on macOS 14.x

• Commit ID: 241ad54
• Duration 1:23:26
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr on Linux RHEL 9

• Commit ID: 241ad54
• Duration 1:25:44
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr-clang on Linux RHEL 9

• Commit ID: 241ad54
• Duration 1:29:14
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr-cluster-tests on Linux RHEL 9

• Commit ID: 241ad54
• Duration 1:42:01
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)
• Cluster Test Logs zip file of the test logs (available for 30 days)

[foundationdb-ci]

Result of foundationdb-pr-macos-m1 on macOS 14.x

• Commit ID: 241ad54
• Duration 2:58:53
• Result: ✅ SUCCEEDED
• Error: N/A
• Build Log terminal output (available for 30 days)
• Build Workspace zip file of the working directory (available for 30 days)