Skip to content

roadrunner: update to v0.16.2 - #1757

Merged
MDA2AV merged 7 commits into
MDA2AV:mainfrom
williamthome:entries/roadrunner-0.16.0
Nov 13, 2026
Merged

MDA2AV merged 7 commits into
MDA2AV:mainfrom
williamthome:entries/roadrunner-0.16.0

Conversation

@williamthome

@williamthome williamthome commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Description

Updates roadrunner from v0.13.1 to v0.16.2, replacing #1751 (v0.15.0). From 0.14, roadrunner accepts new connections without waiting for each connection process to start, so many clients connecting at once to a busy server are accepted sooner; WebSocket compression is off by default; and WebSocket sessions read with a buffer that grows only for large frames, so the entry drops its fixed 4 KB ws.recv_buffer and uses the default. From 0.15, plain-TCP connections wait with a 16 KB read buffer instead of 64 KB and grow it only when a large body arrives, and QUIC sockets no longer set reuseaddr, so an HTTP/3 port is never shared with another socket. From 0.16, a connection keeps the heap its last request needed as the size the next one starts at (HTTP/2 and HTTP/3 stream workers start at the size their connection learned, and WebSocket sessions learn it from handle_info/2), so a handler building a sizeable response, such as the JSON ones, garbage-collects once or twice per request instead of several times; it is on by default and the entry sets nothing for it. In 0.16.1, plain-TCP connections and WebSocket sessions start with a 4 KB read buffer instead of 16 KB and grow it for large heads, bodies and frames, so idle connections hold less memory; the entry sets nothing for this either. In 0.16.2, WebSocket sessions measure each outgoing payload once instead of twice. No option the entry sets changed meaning. The roadrunner-gateway entry builds from frameworks/roadrunner, so it picks up the new version too.


PR Commands — comment on this PR to trigger (requires collaborator approval):

Command Description
/benchmark -f <framework> Run every test the framework subscribes to
/benchmark -f <framework> -t <test> Run one test only
/benchmark -f <framework> --save Run and save results (updates the leaderboard on merge)
/benchmark -f <framework> -t <test> --save Run one test and save results
/benchmark -f <framework> --compare <other> Measure the deltas against another framework instead of this one
/benchmark-multiple -f <fw1>,<fw2>,... Benchmark several frameworks in one run — takes -t and --save too; saved results land in a single commit
/benchmark-multiple --save No -f needed: benchmark and save every framework the PR touches
/benchmark-test -t <test> Benchmark all enabled frameworks subscribed to <test> and save the results

For /benchmark, always specify -f <framework>; the flags combine in any order. Results come back as a comment with a per-profile table of RPS, p99, CPU and memory — one table per framework on multi runs. A new benchmark comment while a run is in flight queues behind it (one deep) instead of cancelling it. For multi-framework PRs (dependency bumps, same-language refactors) prefer /benchmark-multiple, which runs everything in a single job and commits all saved results together, so no run overwrites another. --compare works on single-framework runs only.

What the deltas are measured against. By default, this framework's own results published on main - answering "did this change help?". When you are tuning a variant or a successor entry, --compare re-bases them on another entry instead:

/benchmark -f genhttp --compare genhttp-kestrel

The reply states which baseline it used, and profiles the other framework does not run show n/a rather than a delta.


Run benchmarks locally

You can validate and benchmark your framework locally with the lite script — no CPU pinning, fixed connection counts, all load generators run in Docker.

./scripts/validate.sh <framework>
./scripts/benchmark-lite.sh <framework> baseline
./scripts/benchmark-lite.sh --load-threads 4 <framework>

Requirements: Docker Engine on Linux. Load generators (gcannon, h2load, h2load-h3, wrk) are built as self-contained Docker images on first run.

@williamthome

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f roadrunner,roadrunner-gateway

@github-actions

Copy link
Copy Markdown
Contributor

👋 Benchmark request received. A collaborator will review and approve the run.

@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: all tests

✅ roadrunner

Test Conn RPS Rate CPU/req p99 p99.9 CPU Mem Δ RPS Δ Mem
baseline 4096 1,192,703 — — — — 6702.2% 400MiB +0.3% +11.1%
pipelined 4096 6,297,139 — — — — 6279.6% 310MiB +8.2% +28.1%
limited-conn 4096 524,946 — — — — 3644.5% 267MiB +1.2% +15.1%
json-comp 4096 147,566 — — — — 6298.1% 934MiB +32.8% +45.7%
json-comp 16384 148,188 — — — — 6183.4% 958MiB +41.5% +38.2%
json-tls 4096 224,141 — — — — 6201.6% 1.9GiB +35.3% +125.4%
8gbit 512 48,373 0.9675 ⚠️ 288.76us 998.0us 7258.0us 1337.7% 486MiB -0.2% +2.3%
static-tls 1024 86,659 — — — — 6032.5% 720MiB +3.9% +1.3%
async-db 1024 160,227 — — — — 6414.7% 1.1GiB +30.6% +55.6%
fortunes 1024 70,329 — — — — 6453.0% 1.1GiB -0.3% +23.1%
baseline-h2 256 607,689 — — — — 6557.6% 354MiB +0.3% -1.7%
baseline-h2 1024 609,592 — — — — 6513.2% 465MiB +0.3% -2.9%
static-h2 256 74,610 — — — — 5857.4% 437MiB -6.0% -17.2%
static-h2 1024 72,820 — — — — 6197.0% 1.0GiB -3.5% ~0%
baseline-h2c 256 953,992 — — — — 6600.9% 251MiB -0.9% -4.6%
baseline-h2c 1024 933,964 — — — — 6558.0% 387MiB -1.6% +4.3%
baseline-h2c 4096 847,926 — — — — 6322.3% 605MiB -2.4% -4.1%
json-h2c 1024 136,323 — — — — 3303.9% 1.0GiB +23.7% +14.4%
json-h2c 4096 139,769 — — — — 3986.4% 2.2GiB +35.6% +69.2%
baseline-h3 64 186,051 — — — — 3917.8% 354MiB +1.0% +9.9%
static-h3 64 41,022 — — — — 2897.1% 1.4GiB +9.3% +7.7%
echo-ws 512 1,949,099 — — — — 6794.0% 194MiB -1.0% -0.5%
echo-ws 4096 1,914,161 — — — — 6436.2% 281MiB -2.4% +5.6%
echo-ws 16384 1,745,998 — — — — 6253.2% 417MiB -6.3% +21.9%
echo-ws-pipeline 512 17,670,908 — — — — 6654.9% 199MiB -0.7% +4.7%
echo-ws-pipeline 4096 16,935,305 — — — — 6397.1% 300MiB +1.0% -1.3%
echo-ws-pipeline 16384 16,024,870 — — — — 6585.2% 396MiB -2.3% +20.4%
echo-ws-limited 512 540,383 — — — — 3304.3% 208MiB +0.6% -12.6%
echo-ws-limited 4096 528,179 — — — — 3343.0% 212MiB -1.8% -21.5%
latency-1m 1024 997,425 0.9974 62.42us 21032.0us 36304.0us 6465.6% 457MiB ~0% -5.2%
latency-10k 1024 9,981 0.9981 119.32us 142.0us 781.0us 115.6% 197MiB ~0% +2.1%
latency-500k-8cpu 1024 178,004 0.3560 ⚠️ 44.82us 12849152.0us 13103104.0us 774.2% 146MiB +1.6% ~0%
async 32000 1,141,605 — — — — 6341.3% 1.0GiB -0.1% +13.9%

Composite score, main against this run, in the board's default view:

Family main this run Δ
HTTP/1.1 2566.7 (58 of 107) 2666.5 (57 of 107) +99.8
HTTP/2 273.1 (34 of 40) 282.0 (34 of 40) +8.9
HTTP/3 166.1 (17 of 19) 178.6 (17 of 19) +12.4
Gateway 1834.7 (5 of 9) 1804.7 (5 of 9) -30.0
WebSocket 1018.0 (23 of 38) 999.2 (23 of 38) -18.8

The results on main were measured in another session, so a few points up or down can be the machine and not the change.

✅ roadrunner-gateway

Test Conn RPS CPU Mem Δ RPS Δ Mem
gateway-64 512 157,440 5988.4% 4.5GiB +2.1% +55.2%
gateway-64 1024 151,470 5574.9% 6.4GiB +0.6% +39.1%
gateway-h3 64 79,789 5248.0% 904MiB -1.8% +15.5%
gateway-h3 256 58,242 4069.8% 2.1GiB +0.2% +23.5%
production-stack 256 222,057 5287.3% 1.5GiB -5.9% ~0%
production-stack 1024 193,520 5459.2% 2.1GiB -0.9% ~0%

@williamthome

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f roadrunner,roadrunner-gateway --save

@github-actions

Copy link
Copy Markdown
Contributor

👋 Benchmark request received. A collaborator will review and approve the run.

@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: all tests

✅ roadrunner

Test Conn RPS Rate CPU/req p99 p99.9 CPU Mem Δ RPS Δ Mem
baseline 4096 1,138,679 — — — — 6711.2% 415MiB -4.3% +15.3%
pipelined 4096 5,672,619 — — — — 6261.9% 312MiB -2.6% +28.9%
limited-conn 4096 510,823 — — — — 3772.7% 269MiB -1.5% +15.9%
json-comp 4096 145,867 — — — — 6481.2% 954MiB +31.3% +48.8%
json-comp 16384 142,049 — — — — 6169.6% 949MiB +35.6% +36.9%
json-tls 4096 226,716 — — — — 6196.2% 1.7GiB +36.8% +101.7%
8gbit 512 48,483 0.9697 ⚠️ 289.87us 613.0us 7554.0us 1321.9% 474MiB ~0% -0.2%
static-tls 1024 87,268 — — — — 6082.7% 709MiB +4.6% -0.3%
async-db 1024 163,859 — — — — 6374.1% 1.0GiB +33.5% +41.4%
fortunes 1024 69,879 — — — — 6290.6% 1.2GiB -0.9% +34.3%
baseline-h2 256 597,551 — — — — 6570.2% 347MiB -1.4% -3.6%
baseline-h2 1024 598,371 — — — — 6600.9% 471MiB -1.6% -1.7%
static-h2 256 80,212 — — — — 6140.7% 508MiB +1.0% -3.8%
static-h2 1024 72,732 — — — — 6160.3% 1.0GiB -3.6% ~0%
baseline-h2c 256 936,244 — — — — 6673.0% 255MiB -2.7% -3.0%
baseline-h2c 1024 917,156 — — — — 6498.2% 386MiB -3.4% +4.0%
baseline-h2c 4096 835,437 — — — — 6333.2% 636MiB -3.9% +0.8%
json-h2c 1024 137,515 — — — — 3044.5% 1.1GiB +24.8% +25.9%
json-h2c 4096 139,716 — — — — 4039.8% 2.1GiB +35.5% +61.5%
baseline-h3 64 182,387 — — — — 3759.5% 334MiB -1.0% +3.7%
static-h3 64 42,585 — — — — 3795.6% 1.4GiB +13.5% +7.7%
echo-ws 512 1,872,425 — — — — 6707.4% 194MiB -4.9% -0.5%
echo-ws 4096 1,878,319 — — — — 6633.2% 290MiB -4.2% +9.0%
echo-ws 16384 1,720,550 — — — — 6147.6% 426MiB -7.7% +24.6%
echo-ws-pipeline 512 17,272,726 — — — — 6594.8% 195MiB -2.9% +2.6%
echo-ws-pipeline 4096 17,085,494 — — — — 6438.0% 295MiB +1.9% -3.0%
echo-ws-pipeline 16384 16,173,676 — — — — 6569.6% 384MiB -1.4% +16.7%
echo-ws-limited 512 529,719 — — — — 3308.3% 198MiB -1.4% -16.8%
echo-ws-limited 4096 533,540 — — — — 3418.5% 210MiB -0.8% -22.2%
latency-1m 1024 998,206 0.9982 62.89us 23320.0us 41296.0us 6557.8% 419MiB ~0% -13.1%
latency-10k 1024 9,980 0.9980 108.35us 157.0us 7698.0us 99.3% 197MiB ~0% +2.1%
latency-500k-8cpu 1024 175,818 0.3516 ⚠️ 45.39us 13021184.0us 13283328.0us 772.8% 145MiB +0.3% -0.7%
async 32000 1,124,751 — — — — 6332.6% 1.0GiB -1.6% +13.9%

Composite score, main against this run, in the board's default view:

Family main this run Δ
HTTP/1.1 2566.7 (58 of 107) 2644.7 (57 of 107) +78.0
HTTP/2 273.1 (34 of 40) 282.8 (34 of 40) +9.6
HTTP/3 166.1 (17 of 19) 183.2 (17 of 19) +17.1
Gateway 1834.7 (5 of 9) 1807.7 (5 of 9) -27.0
WebSocket 1018.0 (23 of 38) 986.0 (23 of 38) -32.0

The results on main were measured in another session, so a few points up or down can be the machine and not the change.

✅ roadrunner-gateway

Test Conn RPS CPU Mem Δ RPS Δ Mem
gateway-64 512 155,804 6013.5% 4.6GiB +1.0% +58.6%
gateway-64 1024 151,935 5640.8% 6.8GiB +0.9% +47.8%
gateway-h3 64 79,716 5221.6% 861MiB -1.9% +10.0%
gateway-h3 256 57,715 4414.2% 2.2GiB -0.7% +29.4%
production-stack 256 226,919 5595.3% 1.5GiB -3.9% ~0%
production-stack 1024 191,735 5519.9% 2.2GiB -1.8% +4.8%

@williamthome williamthome changed the title roadrunner: update to v0.16.0 roadrunner: update to v0.16.1 Nov 12, 2026
@williamthome

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f roadrunner,roadrunner-gateway

@github-actions

Copy link
Copy Markdown
Contributor

👋 Benchmark request received. A collaborator will review and approve the run.

@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: all tests

✅ roadrunner

Test Conn RPS Rate CPU/req p99 p99.9 CPU Mem Δ RPS Δ Mem
baseline 4096 1,188,348 — — — — 6753.0% 392MiB ~0% +8.9%
pipelined 4096 6,533,524 — — — — 6287.1% 312MiB +12.2% +28.9%
limited-conn 4096 522,337 — — — — 3694.4% 286MiB +0.7% +23.3%
json-comp 4096 148,913 — — — — 6282.1% 880MiB +34.0% +37.3%
json-comp 16384 152,497 — — — — 6134.6% 879MiB +45.6% +26.8%
json-tls 4096 227,447 — — — — 6205.4% 1.8GiB +37.3% +113.6%
8gbit 512 48,487 0.9697 ⚠️ 277.89us 778.0us 6198.0us 1289.6% 476MiB ~0% +0.2%
static-tls 1024 88,454 — — — — 5816.5% 714MiB +6.0% +0.4%
async-db 1024 166,908 — — — — 6459.4% 983MiB +36.0% +35.8%
fortunes 1024 70,328 — — — — 6431.2% 1.1GiB -0.3% +23.1%
baseline-h2 256 610,301 — — — — 6664.6% 352MiB +0.7% -2.2%
baseline-h2 1024 607,057 — — — — 6520.8% 456MiB -0.1% -4.8%
static-h2 256 86,913 — — — — 5726.0% 506MiB +9.5% -4.2%
static-h2 1024 74,361 — — — — 6135.5% 980MiB -1.4% -4.3%
baseline-h2c 256 945,129 — — — — 6607.8% 258MiB -1.8% -1.9%
baseline-h2c 1024 931,591 — — — — 6554.9% 329MiB -1.8% -11.3%
baseline-h2c 4096 854,567 — — — — 6482.1% 611MiB -1.7% -3.2%
json-h2c 1024 137,933 — — — — 3219.4% 1001MiB +25.2% +11.8%
json-h2c 4096 140,715 — — — — 3969.9% 2.1GiB +36.5% +61.5%
baseline-h3 64 186,874 — — — — 3810.2% 350MiB +1.5% +8.7%
static-h3 64 40,499 — — — — 3041.7% 1.2GiB +7.9% -7.7%
echo-ws 512 1,966,849 — — — — 6788.8% 192MiB -0.1% -1.5%
echo-ws 4096 1,905,289 — — — — 6459.2% 276MiB -2.8% +3.8%
echo-ws 16384 1,759,692 — — — — 6341.0% 422MiB -5.6% +23.4%
echo-ws-pipeline 512 17,752,876 — — — — 6595.2% 234MiB -0.2% +23.2%
echo-ws-pipeline 4096 17,244,377 — — — — 6641.3% 284MiB +2.8% -6.6%
echo-ws-pipeline 16384 16,146,211 — — — — 6153.7% 429MiB -1.6% +30.4%
echo-ws-limited 512 542,279 — — — — 3292.4% 194MiB +1.0% -18.5%
echo-ws-limited 4096 538,079 — — — — 3192.7% 198MiB ~0% -26.7%
latency-1m 1024 997,623 0.9976 62.51us 20280.0us 37264.0us 6485.8% 378MiB ~0% -21.6%
latency-10k 1024 9,983 0.9983 127.84us 140.0us 466.0us 124.3% 195MiB ~0% +1.0%
latency-500k-8cpu 1024 176,240 0.3525 ⚠️ 45.29us 12914688.0us 13152256.0us 773.0% 140MiB +0.6% -4.1%
async 32000 1,128,800 — — — — 6362.6% 1019MiB -1.3% +13.3%

Composite score, main against this run, in the board's default view:

Family main this run Δ
HTTP/1.1 2566.7 (58 of 107) 2685.0 (57 of 107) +118.3
HTTP/2 273.1 (34 of 40) 289.8 (34 of 40) +16.7
HTTP/3 166.1 (17 of 19) 176.9 (17 of 19) +10.8
Gateway 1834.7 (5 of 9) 1850.7 (5 of 9) +16.0
WebSocket 1018.0 (23 of 38) 1006.7 (23 of 38) -11.3

The results on main were measured in another session, so a few points up or down can be the machine and not the change.

✅ roadrunner-gateway

Test Conn RPS CPU Mem Δ RPS Δ Mem
gateway-64 512 155,948 5762.4% 4.5GiB +1.1% +55.2%
gateway-64 1024 150,842 5919.2% 6.4GiB +0.2% +39.1%
gateway-h3 64 80,681 5158.7% 873MiB -0.7% +11.5%
gateway-h3 256 64,988 4645.3% 1.4GiB +11.8% -17.6%
production-stack 256 237,009 5687.8% 1.5GiB +0.4% ~0%
production-stack 1024 192,406 5533.2% 2.1GiB -1.4% ~0%

@williamthome

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f roadrunner,roadrunner-gateway --save

@github-actions

Copy link
Copy Markdown
Contributor

👋 Benchmark request received. A collaborator will review and approve the run.

@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: all tests | Failed: 1

✅ roadrunner

Test Conn RPS Rate CPU/req p99 p99.9 CPU Mem Δ RPS Δ Mem
baseline 4096 1,169,909 — — — — 6699.8% 408MiB -1.6% +13.3%
pipelined 4096 5,543,115 — — — — 6269.3% 254MiB -4.8% +5.0%
limited-conn 4096 518,569 — — — — 3996.9% 301MiB ~0% +29.7%
json-comp 4096 152,700 — — — — 6438.1% 903MiB +37.4% +40.9%
json-comp 16384 141,531 — — — — 6429.9% 951MiB +35.1% +37.2%
json-tls 4096 228,614 — — — — 6186.8% 1.9GiB +38.0% +125.4%
8gbit 512 48,415 0.9683 ⚠️ 282.26us 659.0us 5002.0us 1261.2% 466MiB -0.1% -1.9%
static-tls 1024 88,349 — — — — 5683.1% 708MiB +5.9% -0.4%
async-db 1024 165,371 — — — — 6346.9% 1.1GiB +34.7% +55.6%
fortunes 1024 69,824 — — — — 6429.8% 1014MiB -1.0% +10.8%
baseline-h2 256 598,542 — — — — 6570.7% 379MiB -1.2% +5.3%
baseline-h2 1024 601,869 — — — — 6636.9% 463MiB -1.0% -3.3%
static-h2 256 87,862 — — — — 5620.2% 509MiB +10.7% -3.6%
static-h2 1024 74,046 — — — — 6122.9% 1.0GiB -1.9% ~0%
baseline-h2c 256 933,337 — — — — 6676.1% 258MiB -3.1% -1.9%
baseline-h2c 1024 912,638 — — — — 6538.8% 346MiB -3.8% -6.7%
baseline-h2c 4096 840,930 — — — — 6296.8% 590MiB -3.2% -6.5%
json-h2c 1024 138,330 — — — — 3236.5% 985MiB +25.5% +10.1%
json-h2c 4096 140,668 — — — — 3961.0% 2.2GiB +36.4% +69.2%
baseline-h3 64 184,409 — — — — 3857.1% 335MiB +0.1% +4.0%
static-h3 64 54,088 — — — — 4512.7% 1.3GiB +44.1% ~0%
echo-ws 512 1,890,209 — — — — 6757.8% 195MiB -4.0% ~0%
echo-ws 4096 1,919,907 — — — — 6749.9% 250MiB -2.1% -6.0%
echo-ws 16384 1,715,824 — — — — 6640.2% 410MiB -7.9% +19.9%
echo-ws-pipeline 512 17,308,262 — — — — 6745.9% 188MiB -2.7% -1.1%
echo-ws-pipeline 4096 16,730,243 — — — — 6690.1% 295MiB -0.3% -3.0%
echo-ws-pipeline 16384 15,660,924 — — — — 6325.1% 430MiB -4.6% +30.7%
echo-ws-limited 512 530,607 — — — — 3256.8% 195MiB -1.2% -18.1%
echo-ws-limited 4096 531,405 — — — — 3345.3% 202MiB -1.2% -25.2%
latency-1m 1024 997,158 0.9972 62.78us 25976.0us 55408.0us 6503.9% 353MiB ~0% -26.8%
latency-10k 1024 9,981 0.9981 128.58us 135.0us 1384.0us 124.7% 207MiB ~0% +7.3%
latency-500k-8cpu 1024 173,131 0.3463 ⚠️ 46.08us 13012992.0us 13242368.0us 771.3% 137MiB -1.2% -6.2%
async 32000 1,115,540 — — — — 6354.7% 1.0GiB -2.4% +13.9%

Composite score, main against this run, in the board's default view:

Family main this run Δ
HTTP/1.1 2566.7 (58 of 107) 2660.4 (57 of 107) +93.7
HTTP/2 273.1 (34 of 40) 288.1 (34 of 40) +15.0
HTTP/3 166.1 (17 of 19) 223.4 (17 of 19) +57.2
Gateway 1834.7 (5 of 9) 1807.7 (5 of 9) -27.0
WebSocket 1018.0 (23 of 38) 985.2 (23 of 38) -32.7

The results on main were measured in another session, so a few points up or down can be the machine and not the change.

❌ roadrunner-gateway

Test Conn RPS CPU Mem Δ RPS Δ Mem
gateway-64 512 154,940 5953.9% 4.7GiB +0.5% +62.1%
gateway-64 1024 151,301 5651.9% 6.6GiB +0.5% +43.5%
gateway-h3 64 80,503 5131.4% 913MiB -0.9% +16.6%
gateway-h3 256 57,106 4048.9% 2.1GiB -1.7% +23.5%
Log tail
#41 DONE 0.2s

#43 [authsvc] resolving provenance for metadata file
#43 DONE 0.0s
 Image httparena-roadrunner-gateway-production-stack-authsvc Built 
 Image httparena-roadrunner-gateway-production-stack-server Built 
 Image httparena-roadrunner-gateway-production-stack-edge Built 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Created 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Created 
 Container httparena-roadrunner-gateway-production-stack-server-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-server-1 Created 
 Container httparena-roadrunner-gateway-production-stack-edge-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-edge-1 Created 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Started 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Started 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Healthy 
 Container httparena-roadrunner-gateway-production-stack-server-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-server-1 Started 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-server-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Error dependency authsvc failed to start
dependency failed to start: container httparena-roadrunner-gateway-production-stack-authsvc-1 exited (0)

─── httparena-roadrunner-gateway-production-stack-edge-1 — status=created exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-edge-1 — the container produced no output at all

─── httparena-roadrunner-gateway-production-stack-server-1 — status=running exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-server-1 — last 120 log lines ───
  | Exec: /app/erts-17.0.2/bin/erlexec -noinput +Bd -boot /app/releases/0.1.0/start -mode embedded -boot_var SYSTEM_LIB_DIR /app/lib -config /app/releases/0.1.0/sys.config -args_file /app/releases/0.1.0/vm.args -- foreground
  | Root: /app
  | /app
  | =ESOCK WARNING MSG==== 13-Nov-2026::08:09:06.279486 ===

  | [UNIX-ESSIO] Failed open sctp dynamic library: libsctp.so.1

─── httparena-roadrunner-gateway-production-stack-server-1 — end of logs ───

─── httparena-roadrunner-gateway-production-stack-cache-1 — status=running exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-cache-1 — last 120 log lines ───
  | 7:C 13 Nov 2026 08:09:04.420 # WARNING Memory overcommit must be enabled! Without it, a background save or replication may fail under low memory condition. Being disabled, it can also cause failures without low memory condition, see https://github.com/jemalloc/jemalloc/issues/1328. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.
  | 7:C 13 Nov 2026 08:09:04.420 * oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
  | 7:C 13 Nov 2026 08:09:04.420 * Redis version=7.4.8, bits=64, commit=00000000, modified=0, pid=7, just started
  | 7:C 13 Nov 2026 08:09:04.420 * Configuration loaded
  | 7:M 13 Nov 2026 08:09:04.420 * monotonic clock: POSIX clock_gettime
  | 7:M 13 Nov 2026 08:09:04.424 * Running mode=standalone, port=6379.
  | 7:M 13 Nov 2026 08:09:04.424 * Server initialized
  | 7:M 13 Nov 2026 08:09:04.425 * Ready to accept connections tcp
  | [redis] seeded from /seed.txt
─── httparena-roadrunner-gateway-production-stack-cache-1 — end of logs ───

─── httparena-roadrunner-gateway-production-stack-authsvc-1 — status=exited exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-authsvc-1 — the container produced no output at all
[FAIL] gateway compose up failed
httparena-postgres
[info] restoring loopback MTU to 65536

@williamthome williamthome changed the title roadrunner: update to v0.16.1 roadrunner: update to v0.16.2 Nov 13, 2026
@williamthome

Copy link
Copy Markdown
Contributor Author

/benchmark-multiple -f roadrunner,roadrunner-gateway --save

@github-actions

Copy link
Copy Markdown
Contributor

👋 Benchmark request received. A collaborator will review and approve the run.

@github-actions

Copy link
Copy Markdown
Contributor

Benchmark Results

Frameworks: 2 | Test: all tests | Failed: 1

✅ roadrunner

Test Conn RPS Rate CPU/req p99 p99.9 CPU Mem Δ RPS Δ Mem
baseline 4096 1,190,380 — — — — 6409.9% 382MiB ~0% +6.1%
pipelined 4096 6,106,728 — — — — 6289.3% 329MiB +4.9% +36.0%
limited-conn 4096 530,359 — — — — 3621.4% 281MiB +2.2% +21.1%
json-comp 4096 150,319 — — — — 6340.3% 891MiB +35.3% +39.0%
json-comp 16384 148,211 — — — — 6492.6% 923MiB +41.5% +33.2%
json-tls 4096 223,898 — — — — 6158.5% 1.8GiB +35.1% +113.6%
8gbit 512 48,426 0.9685 ⚠️ 276.68us 847.0us 5666.0us 1240.5% 462MiB -0.1% -2.7%
static-tls 1024 89,655 — — — — 5663.6% 721MiB +7.4% +1.4%
async-db 1024 165,313 — — — — 6379.1% 1.0GiB +34.7% +41.4%
fortunes 1024 69,357 — — — — 6424.1% 1.4GiB -1.7% +56.7%
baseline-h2 256 608,773 — — — — 6529.8% 423MiB +0.5% +17.5%
baseline-h2 1024 601,261 — — — — 6513.9% 478MiB -1.1% -0.2%
static-h2 256 73,012 — — — — 6278.1% 478MiB -8.0% -9.5%
static-h2 1024 74,934 — — — — 5910.4% 998MiB -0.7% -2.5%
baseline-h2c 256 942,518 — — — — 6654.7% 257MiB -2.1% -2.3%
baseline-h2c 1024 925,316 — — — — 6506.3% 346MiB -2.5% -6.7%
baseline-h2c 4096 853,129 — — — — 6271.0% 578MiB -1.8% -8.4%
json-h2c 1024 135,561 — — — — 3230.1% 1.1GiB +23.0% +25.9%
json-h2c 4096 140,766 — — — — 3867.1% 2.1GiB +36.5% +61.5%
baseline-h3 64 180,097 — — — — 3903.1% 352MiB -2.2% +9.3%
static-h3 64 43,952 — — — — 3879.1% 1.4GiB +17.1% +7.7%
echo-ws 512 1,944,473 — — — — 6621.1% 189MiB -1.3% -3.1%
echo-ws 4096 1,889,351 — — — — 6498.0% 246MiB -3.6% -7.5%
echo-ws 16384 1,755,757 — — — — 6256.9% 404MiB -5.8% +18.1%
echo-ws-pipeline 512 17,175,317 — — — — 6563.7% 237MiB -3.5% +24.7%
echo-ws-pipeline 4096 17,633,724 — — — — 6726.6% 306MiB +5.1% +0.7%
echo-ws-pipeline 16384 16,224,509 — — — — 6252.8% 422MiB -1.1% +28.3%
echo-ws-limited 512 539,392 — — — — 3314.5% 191MiB +0.4% -19.7%
echo-ws-limited 4096 533,500 — — — — 3303.3% 195MiB -0.8% -27.8%
latency-1m 1024 997,716 0.9977 62.46us 19496.0us 35312.0us 6479.1% 376MiB ~0% -22.0%
latency-10k 1024 9,981 0.9981 132.93us 140.0us 16968.0us 128.1% 187MiB ~0% -3.1%
latency-500k-8cpu 1024 171,678 0.3434 ⚠️ 46.49us 13111296.0us 13398016.0us 772.0% 135MiB -2.0% -7.5%
async 32000 1,133,425 — — — — 6393.6% 1002MiB -0.9% +11.5%

Composite score, main against this run, in the board's default view:

Family main this run Δ
HTTP/1.1 2566.7 (58 of 107) 2658.4 (57 of 107) +91.7
HTTP/2 273.1 (34 of 40) 281.5 (34 of 40) +8.4
HTTP/3 166.1 (17 of 19) 187.5 (17 of 19) +21.4
Gateway 1834.7 (5 of 9) 1807.7 (5 of 9) -27.0
WebSocket 1018.0 (23 of 38) 1000.9 (23 of 38) -17.1

The results on main were measured in another session, so a few points up or down can be the machine and not the change.

❌ roadrunner-gateway

Test Conn RPS CPU Mem Δ RPS Δ Mem
gateway-64 512 156,035 5985.3% 4.5GiB +1.2% +55.2%
gateway-64 1024 151,330 5802.2% 6.4GiB +0.5% +39.1%
gateway-h3 64 79,929 5243.2% 869MiB -1.6% +11.0%
gateway-h3 256 58,808 4093.8% 2.1GiB +1.2% +23.5%
Log tail
#42 DONE 0.1s

#43 [authsvc] resolving provenance for metadata file
#43 DONE 0.0s
 Image httparena-roadrunner-gateway-production-stack-server Built 
 Image httparena-roadrunner-gateway-production-stack-authsvc Built 
 Image httparena-roadrunner-gateway-production-stack-edge Built 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Created 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Created 
 Container httparena-roadrunner-gateway-production-stack-server-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-server-1 Created 
 Container httparena-roadrunner-gateway-production-stack-edge-1 Creating 
 Container httparena-roadrunner-gateway-production-stack-edge-1 Created 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Started 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Started 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-cache-1 Healthy 
 Container httparena-roadrunner-gateway-production-stack-server-1 Starting 
 Container httparena-roadrunner-gateway-production-stack-server-1 Started 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-server-1 Waiting 
 Container httparena-roadrunner-gateway-production-stack-authsvc-1 Error dependency authsvc failed to start
dependency failed to start: container httparena-roadrunner-gateway-production-stack-authsvc-1 exited (0)

─── httparena-roadrunner-gateway-production-stack-edge-1 — status=created exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-edge-1 — the container produced no output at all

─── httparena-roadrunner-gateway-production-stack-server-1 — status=running exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-server-1 — last 120 log lines ───
  | Exec: /app/erts-17.0.2/bin/erlexec -noinput +Bd -boot /app/releases/0.1.0/start -mode embedded -boot_var SYSTEM_LIB_DIR /app/lib -config /app/releases/0.1.0/sys.config -args_file /app/releases/0.1.0/vm.args -- foreground
  | Root: /app
  | /app
  | =ESOCK WARNING MSG==== 13-Nov-2026::14:07:53.138385 ===

  | [UNIX-ESSIO] Failed open sctp dynamic library: libsctp.so.1

─── httparena-roadrunner-gateway-production-stack-server-1 — end of logs ───

─── httparena-roadrunner-gateway-production-stack-cache-1 — status=running exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-cache-1 — last 120 log lines ───
  | 7:C 13 Nov 2026 14:07:51.282 # WARNING Memory overcommit must be enabled! Without it, a background save or replication may fail under low memory condition. Being disabled, it can also cause failures without low memory condition, see https://github.com/jemalloc/jemalloc/issues/1328. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.
  | 7:C 13 Nov 2026 14:07:51.282 * oO0OoO0OoO0Oo Redis is starting oO0OoO0OoO0Oo
  | 7:C 13 Nov 2026 14:07:51.282 * Redis version=7.4.8, bits=64, commit=00000000, modified=0, pid=7, just started
  | 7:C 13 Nov 2026 14:07:51.282 * Configuration loaded
  | 7:M 13 Nov 2026 14:07:51.282 * monotonic clock: POSIX clock_gettime
  | 7:M 13 Nov 2026 14:07:51.285 * Running mode=standalone, port=6379.
  | 7:M 13 Nov 2026 14:07:51.286 * Server initialized
  | 7:M 13 Nov 2026 14:07:51.286 * Ready to accept connections tcp
  | [redis] seeded from /seed.txt
─── httparena-roadrunner-gateway-production-stack-cache-1 — end of logs ───

─── httparena-roadrunner-gateway-production-stack-authsvc-1 — status=exited exit=0 oom=false error=
─── httparena-roadrunner-gateway-production-stack-authsvc-1 — the container produced no output at all
[FAIL] gateway compose up failed
httparena-postgres
[info] restoring loopback MTU to 65536

@williamthome

Copy link
Copy Markdown
Contributor Author

@MDA2AV the roadrunner-gateway production-stack failed on both runs here: the shared authsvc container exits with code 0 without any output, so the stack never comes up (log tail above). The roadrunner results and the other gateway profiles look fine. Could you take a look at authsvc on the bench host when you have a moment? Or is this PR safe to merge as is?

MDA2AV added a commit that referenced this pull request Nov 13, 2026
…uthsvc on Rust 1.100, unbreak the aspnet gateway entries (#1767)

* validate.sh: stop the standalone container before the compose profiles

The standalone container runs on --network host whenever an entry
subscribes to a compose profile, and it was left running while the
gateway and production-stack stacks came up. An entry that also serves
h2/h3 keeps 8443 (and 8080), so the stack's edge failed to bind, shared
the port, or exited on the QUIC port, and the checks reached the
standalone server instead of the stack.

* authsvc: ship the real binary under cargo 1.100's build-dir layout

production-stack has failed on the bench host since the morning of
2026-11-13 on every entry (roadrunner-gateway #1757, morojs #1761,
tako-gateway #1762): the shared authsvc container exits 0 with no output,
the edge never sees it healthy, and the run ends with no results.

rust:1-alpine moved to Rust 1.100 on 2026-11-12. Its cargo keeps each
package's artifacts under target/release/build/<package>/<hash>/ rather
than target/release/deps/, so the cache-priming step's
`rm -rf target/release/deps/authsvc*` stopped matching anything. The
placeholder built from `fn main() {}` stayed up to date, the second
`cargo build` finished in 0.04s without compiling src/, and the image
shipped the placeholder: a 418 KB binary that exits 0 at once.

`cargo clean --release -p authsvc` removes the placeholder's artifacts
wherever cargo keeps them, and a grep for the server's startup line fails
the build if the placeholder is ever shipped again. Built against the
current rust:1-alpine (1.100.0): a 1.2 MB binary that logs
"authsvc listening on ...", answers /_health with 200 and an /_auth
without a token with 401.

* aspnet-minimal_nginx, aspnet-minimal_caddy: build against aspnet-minimal's Services and Types

Both Dockerfiles still COPY aspnet-minimal's AppData.cs and Models.cs,
which #1002 (2026-07-18) split into Services/ and Types/, so neither image
has built since. They were the only failures in #1615's validation sweep,
and their published results have not been refreshed since July.

Copy Handlers.cs with Services/ and Types/, register DatabaseService,
DatasetService, ItemService and the AppJsonContext resolver the way
aspnet-minimal's Program.cs does, and give the nginx entry's /api routes
DatabaseService's Postgres pool in place of AppData.PgDataSource. The
caddy entry gains the StackExchange.Redis reference DatabaseService needs;
it connects to nothing unless REDIS_URL is set, which gateway-h3 never
does.

---------

Co-authored-by: Jacob Rothstein <hi@jbr.me>
@MDA2AV

MDA2AV commented Nov 13, 2026

Copy link
Copy Markdown
Owner

Thanks for flagging it. The problem was the shared authsvc, not roadrunner. rust:1-alpine moved to Rust 1.100 on 11-12, and under that cargo the Dockerfile's dependency-caching step shipped its fn main() {} placeholder instead of the real server. That's why the container exited 0 with no output. It's fixed on main in #1767.

Nothing on this branch needs to change. A production-stack re-run with --save merges main into the branch before building, so it gets the fixed authsvc. A run without --save builds this branch as it is and would fail the same way.

@williamthome

Copy link
Copy Markdown
Contributor Author

Thanks for the quick fix and the explanation! If everything looks good on your side, this one is ready to merge.

@MDA2AV
MDA2AV merged commit 0f29e22 into MDA2AV:main Nov 13, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants