Add Memcached meta-command protocol support (memcache_meta)

[sam-obeid]

Adds a new value 'memcache_meta' to the -P/--protocol flag, enabling benchmarking of the Memcached meta protocol (memcached 1.6+).

Wire format

Every command is wire-terminated by 'mn\r\n' so every response is uniformly terminated by 'MN'. This eliminates write-time/parse-time mode-attribution races under pipelined / mixed-command workloads:

SET:        ms <k> <l> T<e>\r\n<value>\r\n mn\r\n  -> HD\r\n MN\r\n
GET:        mg <k> v\r\n mn\r\n                    -> VA <n>\r\n<data>\r\n MN\r\n
                                                        (or EN MN on miss)
multi-GET:  pipelined 'mg ... v\r\n' + 'mn\r\n'      -> ...VA/EN..., MN

The wire cost is +4 request +4 response bytes per command; at typical value sizes (>= 64B) this is < 1% of bandwidth.

Implementation

The new memcache_meta_protocol class inherits from memcache_text_protocol to reuse the buffer plumbing and the seven assert(0) stubs for operations not supported by either Memcached path (auth, select_db, cluster_slots, wait, arbitrary_command). Only the four wire-touching methods are overridden: write_command_set / write_command_get / write_command_multi_get / parse_response.

Performance touches:

• mg / multi-get write paths use raw evbuffer_add (no vsnprintf in the per-key hot path), matching the existing memcache_text multi-get pattern.
• VA parsing uses strtoul (not sscanf) on the GET hot path.
• Response state machine has a single termination rule (read until MN), removing all conditional branches on external state in the hot loop.

Validation against memcached 1.6.41 (4 clients x 2 threads x 10K ops):
meta SETs 135K ops/sec (text: 131K)
meta GETs 130K ops/sec (text: 133K, ~3.5x less response bandwidth)
meta multi-get(4) 256K keys/sec
mixed --pipeline=4 246K ops/sec

Tests

13 unit tests in tests/test_memcache_meta_protocol.py exercise the parser via an in-process mock TCP server that speaks the meta protocol, so the suite runs without requiring memcached or Redis. Coverage:

Happy path: single SET, single GET (hit / miss), GET against
pre-loaded data, multi-get with mn terminator,
SET -> GET roundtrip
Status codes: NS / EX / NF responses do not error
Async / partial: fragmented VA delivery (header and value across
two TCP sends -> rs_read_value 'need more data' path)
Error path: unknown server response classified as parse error
Pipelining: --pipeline=4 SET, mixed --pipeline=2 SET +
multi-get (regression for write-time/parse-time race)
Multi-get: deterministic VA + EN mix within a single pipeline

Tests are also discoverable by RLTest via top-level test_meta_* adapters so they ride along with ./tests/run_tests.sh.

Documentation

Updated --help text, error message for invalid -P, manpage (memtier_benchmark.1), README.md feature bullet, and bash completion.

---

Note

Medium Risk
Adds a new Memcached meta protocol implementation with custom request/response parsing and pipelining behavior, which can impact benchmark correctness and connection stability. Risk is moderated by extensive new protocol-specific integration tests and mainly additive changes.

Overview
Adds support for benchmarking Memcached’s modern meta-command ASCII protocol via a new -P/--protocol=memcache_meta option, including a dedicated protocol implementation (memcache_meta_protocol) and factory/CLI wiring.

Implements meta wire formats for SET/GET/multi-GET and a new response parser that treats MN as a universal terminator to avoid mixed-command pipelining mode races, and updates validation to disallow --authenticate for this protocol.

Updates user-facing docs/help (README, manpage, --help text, bash completion) and adds a comprehensive Python test suite with an in-process mock meta server to verify exact wire format, parsing behavior (hits/misses, fragmentation), error handling, and pipelined mixed workloads.

^{Reviewed by Cursor Bugbot for commit 7d2eddd. Bugbot is set up for automated code reviews on this repo. Configure here.}

[jit-ci]

Hi, I’m Jit, a friendly security platform designed to help developers build secure applications from day zero with an MVS (Minimal viable security) mindset.

In case there are security findings, they will be communicated to you as a comment inside the PR.

Hope you’ll enjoy using Jit.

Questions? Comments? Want to learn more? Get in touch with us.

[sam-obeid]

cursor review

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 7d2eddd. Configure here.}