webapp/srs

Fork 0

mirror of https://github.com/ossrs/srs.git synced 2026-10-24 09:59:20 +03:00

Commit Graph

Author	SHA1	Message	Date
Winlin	7b737e64df	Perf: Load and profile tools, and the first SRS and Go proxy fixes. v8.0.63 (#4785 ) SRS gets performance tools that load and profile SRS in Docker for every protocol: a lean load tool, an internal skill that runs an improvement round, and a measurement method that holds up to a 10% verdict. The same tools then measure the Go proxy. The first runs found one SRS fix and three Go proxy fixes. The load tool, a standalone Go module in cmd/perf (bin/srs-perf): - Per-core worker processes shard the clients and report counters each second. Lean clients for every protocol: RTMP publish and play, HTTP-FLV, HTTP-TS, HLS, DASH, RTSP, SRT, WHIP and WHEP. No decode; frames and lag come from timestamps. - 17 scenarios: each protocol's fan-out, publishers (rtmp-pub, whip-pub), many-streams, cross-protocol bridges (srt-rtmp, srt-whep, whip-rtmp, whip-flv), and delay probes for the RTMP pause, forward and edge. - run ramps by doubling to a capacity, then holds at 50% and 90% of it. compare checks that the setups match and reports capacity and CPU per client at the 10% line. - make perf-image builds an optimized SRS with symbols and frame pointers, with perf, FlameGraph and FFmpeg. SRS runs on one pinned core and is configured only by env files, one per scenario. The measurement method: - A flat CBR source, checked to ±3% per second, since SRS CPU follows the bitrate. - SRS CPU from its cgroup once a second. Holds have no profiler in the measuring window; each count gets its own separate profile hold. - perf-ab.sh builds two revisions and runs them side by side on separate cpusets. compare gives the 95% CI of the paired per-second difference. - perf-verify.sh checks every run: capacity, healthy steps, profiles and symbols, noise, and what bound the run. The internal-perf-for-srs skill: - SKILL.md holds the rules, the shared tools and the round workflow. A Target Router loads references/cpp-server.md for the C++ media server or references/go-server.md for the Go proxy. - Each reference covers its tools, topology, scenarios, measuring, how to read its profiles (Linux perf, or Go pprof and trace), A/B, the verified scenarios with their capacity and top bottleneck, and how to start a round. - srs-develop routes performance, benchmark, load test and profile tasks to it. perf-skill-test.sh checks the text against the scenario registry. - Reports are git-ignored, since their numbers hold only for the machine that made them. - The skills now name each server by its job: the C++ media server and the Go proxy, with no "generation". C++ media server, one core, Linux arm64 in Docker: - Fix: an empty SrsSharedPtr holds no reference count, so each fan-out copy of a message does one malloc instead of three. Covered by srs_utest_ai05.cpp. - Capacity: rtmp-rtmp, rtmp-flv, rtmp-hls, whip-rtmp and whip-flv 12,800; rtmp-dash and srt-rtmp 6,400; rtmp-ts and rtmp-rtsp 3,200; whip-whep 2,000 (cap); rtmp-whep, srt-whep, srt-srt and rtmp-pub 1,280; whip-pub 1,000 (cap); many-streams 640 streams. - Top bottlenecks found, not fixed: the RTMP fan-out copy, _st_epoll_dispatch with EAGAIN reads, 4 KB reads and writes for HLS and DASH segments, one write per message for HTTP-TS and two per packet for RTSP, the SRTP replay window in libsrtp, and libsrt for SRT. - Delays: the paused RTMP player's 500 ms sleep, and the forward's and edge publish's 500 ms receive timeout, each adding about 220 ms of lag. The Go proxy, cmd/proxy, on one core in front of 1 to 4 SRS origins: - perf-up.sh -target proxy, run -target proxy, which samples the proxy and each origin, Go profiles in each profile hold (CPU, allocations, mutex, block, trace, GC), and proxy A/B. - Fix: HTTP-FLV, HTTP-TS and HLS relay the origin's error status, headers and body, rather than a 500 with the origin's address. - Fix: a WebRTC STUN binding routes by its ufrag, not a cached client address. WHIP and WHEP failed from 160 clients on three origins. - Capacity on three origins: rtmp-flv and rtmp-hls 12,800; rtmp-rtmp and rtmp-ts 6,400; whip-whep 2,000 (cap); srt-srt and rtmp-pub 1,280; whip-pub 1,000 (cap); many-streams 1,280 streams. --------- Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-10-21 18:46:30 -04:00
Winlin	326b5fa817	Proxy: Preserve HTTP response headers and safely rewrite HLS. v8.0.20 (#4718 ) ## Summary - Preserve end-to-end backend response headers for HTTP-FLV, HTTP-TS, and HLS responses. - Remove standard hop-by-hop headers and additional fields nominated by `Connection`. - Rewrite HLS segment URLs with a canonical single `&` when the original URL already has query parameters. - Repair response metadata after modifying an m3u8 body, and add wire-level regression coverage. ## Problem The HTTP proxy called `WriteHeader` before copying headers from the backend. Go commits the response at that point, so later additions such as `Content-Type`, `Cache-Control`, `ETag`, `Last-Modified`, `Vary`, and custom origin metadata never reached the downstream client. HLS playlist rewriting also generated segment URLs containing an unnecessary empty query component: ```text segment.ts?spbhid=xxx&&token=abc ``` Moving header copying before `WriteHeader` is not sufficient by itself. An m3u8 playlist is modified when the proxy inserts `spbhid`, so the backend's original `Content-Length`, strong `ETag`, digest, content encoding, and range metadata no longer describe the downstream representation. Forwarding the stale length can make Go reject the enlarged response with `http: wrote more than the declared Content-Length`. A proxy must also avoid forwarding connection-specific headers to the next hop. ## Changes - Copy backend end-to-end headers before committing HTTP-FLV, HTTP-TS, and HLS responses. - Strip `Connection`, its nominated fields, and the standard proxy/connection-specific header set. - Keep byte-transparent FLV and TS representation metadata unchanged. - Read and rewrite m3u8 playlists before sending response headers. - Recompute `Content-Length` and remove stale validators, digests, encodings, and range metadata when the playlist body changes. - Generate existing-query segment URLs as: ```text segment.ts?spbhid=xxx&token=abc ``` - Add real HTTP-boundary tests for header preservation, hop-by-hop removal, query rewriting, complete playlist delivery, and rewritten-body metadata. --------- Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com>	2026-08-12 22:17:39 -04:00
Winlin	6ee6f1ca5f	Proxy: Refactor for testability; add SRT/WHIP E2E and unit tests. v7.0.148 (#4675 ) - Refactor the Go proxy for dependency injection: every proxy server, the bootstrap, the signal handler, the load balancers, and AMF0 now accept functional-option seams (factories/closures) so tests can inject fakes without binding real sockets, talking to real Redis, or racing on package globals. - Drop the package-global `lb.SrsLoadBalancer`. The bootstrap creates the LB locally and threads it through every proxy server constructor. Two old global indirections in `internal/signal` and `internal/rtmp/amf0` are likewise replaced by per-instance fields. - Rename `internal/server` → `internal/proxy` and rename the `lb` public surface for clarity: `SRSLoadBalancer` is split into `OriginService` / `HLSService` / `RTCService` and recomposed as `OriginLoadBalancer`; `SRSServer` → `OriginServer`; all proxy server types gain a `Proxy` qualifier (e.g. `RTMPServer` → `RTMPProxyServer`). - Extract the Redis client behind a new `internal/redisclient` package with a minimal `RedisClient` interface and a counterfeiter fake. - Add counterfeiter fakes (`proxyfakes`, `lbfakes`, `redisclientfakes`) and ~7.5k lines of unit tests covering bootstrap, memory + Redis LBs, all five proxy servers, the signal handler, and AMF0. - Add two new E2E flows — `proxy-e2e-srt-test.sh` (SRT publish through proxy, verify SRT/RTMP/HTTP-FLV/HLS playback) and `proxy-e2e-whip-test.sh` (WHIP publish, verify RTMP/HTTP-FLV/HLS via origin `rtc_to_rtmp`) — plus `setup-ffmpeg-with-whip.sh`, a macOS builder for an ffmpeg with openssl-DTLS WHIP and SRT support that the two scripts auto-invoke when needed. - Workspace reorg: move `memory/` and `skills/` to the repo root so all agent tools (Claude / Codex / Kiro / OpenClaw) share one source of truth via symlinks. Sync `docs/proxy/proxy-load-balancer.md` and `memory/srs-codebase-map.md` with the new names. No protocol, log, HTTP API, or wire-format changes. Refactor only — all externally observable proxy behavior is unchanged. --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com>	2026-05-17 12:09:07 -04:00

Author

SHA1

Message

Date

Winlin

7b737e64df

Perf: Load and profile tools, and the first SRS and Go proxy fixes. v8.0.63 (#4785 )

SRS gets performance tools that load and profile SRS in Docker for every
protocol: a lean load tool, an internal skill that runs an improvement
round, and a measurement method that holds up to a 10% verdict. The same
tools then measure the Go proxy. The first runs found one SRS fix and
three Go proxy fixes.

The load tool, a standalone Go module in cmd/perf (bin/srs-perf):

- Per-core worker processes shard the clients and report counters each
second. Lean clients for every protocol: RTMP publish and play,
HTTP-FLV, HTTP-TS, HLS, DASH, RTSP, SRT, WHIP and WHEP. No decode;
frames and lag come from timestamps.
- 17 scenarios: each protocol's fan-out, publishers (rtmp-pub,
whip-pub), many-streams, cross-protocol bridges (srt-rtmp, srt-whep,
whip-rtmp, whip-flv), and delay probes for the RTMP pause, forward and
edge.
- run ramps by doubling to a capacity, then holds at 50% and 90% of it.
compare checks that the setups match and reports capacity and CPU per
client at the 10% line.
- make perf-image builds an optimized SRS with symbols and frame
pointers, with perf, FlameGraph and FFmpeg. SRS runs on one pinned core
and is configured only by env files, one per scenario.

The measurement method:

- A flat CBR source, checked to ±3% per second, since SRS CPU follows
the bitrate.
- SRS CPU from its cgroup once a second. Holds have no profiler in the
measuring window; each count gets its own separate profile hold.
- perf-ab.sh builds two revisions and runs them side by side on separate
cpusets. compare gives the 95% CI of the paired per-second difference.
- perf-verify.sh checks every run: capacity, healthy steps, profiles and
symbols, noise, and what bound the run.

The internal-perf-for-srs skill:

- SKILL.md holds the rules, the shared tools and the round workflow. A
Target Router loads references/cpp-server.md for the C++ media server or
references/go-server.md for the Go proxy.
- Each reference covers its tools, topology, scenarios, measuring, how
to read its profiles (Linux perf, or Go pprof and trace), A/B, the
verified scenarios with their capacity and top bottleneck, and how to
start a round.
- srs-develop routes performance, benchmark, load test and profile tasks
to it. perf-skill-test.sh checks the text against the scenario registry.
- Reports are git-ignored, since their numbers hold only for the machine
that made them.
- The skills now name each server by its job: the C++ media server and
the Go proxy, with no "generation".

C++ media server, one core, Linux arm64 in Docker:

- Fix: an empty SrsSharedPtr holds no reference count, so each fan-out
copy of a message does one malloc instead of three. Covered by
srs_utest_ai05.cpp.
- Capacity: rtmp-rtmp, rtmp-flv, rtmp-hls, whip-rtmp and whip-flv
12,800; rtmp-dash and srt-rtmp 6,400; rtmp-ts and rtmp-rtsp 3,200;
whip-whep 2,000 (cap); rtmp-whep, srt-whep, srt-srt and rtmp-pub 1,280;
whip-pub 1,000 (cap); many-streams 640 streams.
- Top bottlenecks found, not fixed: the RTMP fan-out copy,
_st_epoll_dispatch with EAGAIN reads, 4 KB reads and writes for HLS and
DASH segments, one write per message for HTTP-TS and two per packet for
RTSP, the SRTP replay window in libsrtp, and libsrt for SRT.
- Delays: the paused RTMP player's 500 ms sleep, and the forward's and
edge publish's 500 ms receive timeout, each adding about 220 ms of lag.

The Go proxy, cmd/proxy, on one core in front of 1 to 4 SRS origins:

- perf-up.sh -target proxy, run -target proxy, which samples the proxy
and each origin, Go profiles in each profile hold (CPU, allocations,
mutex, block, trace, GC), and proxy A/B.
- Fix: HTTP-FLV, HTTP-TS and HLS relay the origin's error status,
headers and body, rather than a 500 with the origin's address.
- Fix: a WebRTC STUN binding routes by its ufrag, not a cached client
address. WHIP and WHEP failed from 160 clients on three origins.
- Capacity on three origins: rtmp-flv and rtmp-hls 12,800; rtmp-rtmp and
rtmp-ts 6,400; whip-whep 2,000 (cap); srt-srt and rtmp-pub 1,280;
whip-pub 1,000 (cap); many-streams 1,280 streams.

---------

Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

2026-10-21 18:46:30 -04:00

Winlin

326b5fa817

Proxy: Preserve HTTP response headers and safely rewrite HLS. v8.0.20 (#4718 )

## Summary

- Preserve end-to-end backend response headers for HTTP-FLV, HTTP-TS,
and HLS responses.
- Remove standard hop-by-hop headers and additional fields nominated by
`Connection`.
- Rewrite HLS segment URLs with a canonical single `&` when the original
URL already has query parameters.
- Repair response metadata after modifying an m3u8 body, and add
wire-level regression coverage.

## Problem

The HTTP proxy called `WriteHeader` before copying headers from the
backend. Go commits the response at that point, so later additions such
as `Content-Type`, `Cache-Control`, `ETag`, `Last-Modified`, `Vary`, and
custom origin metadata never reached the downstream client.

HLS playlist rewriting also generated segment URLs containing an
unnecessary empty query component:

```text
segment.ts?spbhid=xxx&&token=abc
```

Moving header copying before `WriteHeader` is not sufficient by itself.
An m3u8 playlist is modified when the proxy inserts `spbhid`, so the
backend's original `Content-Length`, strong `ETag`, digest, content
encoding, and range metadata no longer describe the downstream
representation. Forwarding the stale length can make Go reject the
enlarged response with `http: wrote more than the declared
Content-Length`.

A proxy must also avoid forwarding connection-specific headers to the
next hop.

## Changes

- Copy backend end-to-end headers before committing HTTP-FLV, HTTP-TS,
and HLS responses.
- Strip `Connection`, its nominated fields, and the standard
proxy/connection-specific header set.
- Keep byte-transparent FLV and TS representation metadata unchanged.
- Read and rewrite m3u8 playlists before sending response headers.
- Recompute `Content-Length` and remove stale validators, digests,
encodings, and range metadata when the playlist body changes.
- Generate existing-query segment URLs as:

  ```text
  segment.ts?spbhid=xxx&token=abc
  ```

- Add real HTTP-boundary tests for header preservation, hop-by-hop
removal, query rewriting, complete playlist delivery, and rewritten-body
metadata.

---------

Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com>

2026-08-12 22:17:39 -04:00

Winlin

6ee6f1ca5f

Proxy: Refactor for testability; add SRT/WHIP E2E and unit tests. v7.0.148 (#4675 )

- Refactor the Go proxy for dependency injection: every proxy server,
the bootstrap, the signal handler, the load balancers, and AMF0 now accept
functional-option seams (factories/closures) so tests can inject fakes
without binding real sockets, talking to real Redis, or racing on
package globals.

- Drop the package-global `lb.SrsLoadBalancer`. The bootstrap creates
the LB locally and threads it through every proxy server constructor. Two old
global indirections in `internal/signal` and `internal/rtmp/amf0` are
likewise replaced by per-instance fields.

- Rename `internal/server` → `internal/proxy` and rename the `lb` public
surface for clarity: `SRSLoadBalancer` is split into `OriginService` /
`HLSService` / `RTCService` and recomposed as `OriginLoadBalancer`;
`SRSServer` → `OriginServer`; all proxy server types gain a `Proxy`
qualifier (e.g. `RTMPServer` → `RTMPProxyServer`).

- Extract the Redis client behind a new `internal/redisclient` package
with a minimal `RedisClient` interface and a counterfeiter fake.

- Add counterfeiter fakes (`proxyfakes`, `lbfakes`, `redisclientfakes`)
and ~7.5k lines of unit tests covering bootstrap, memory + Redis LBs, all
five proxy servers, the signal handler, and AMF0.

- Add two new E2E flows — `proxy-e2e-srt-test.sh` (SRT publish through
proxy, verify SRT/RTMP/HTTP-FLV/HLS playback) and `proxy-e2e-whip-test.sh`
(WHIP publish, verify RTMP/HTTP-FLV/HLS via origin `rtc_to_rtmp`) — plus
`setup-ffmpeg-with-whip.sh`, a macOS builder for an ffmpeg with
openssl-DTLS WHIP and SRT support that the two scripts auto-invoke when needed.

- Workspace reorg: move `memory/` and `skills/` to the repo root so all
agent tools (Claude / Codex / Kiro / OpenClaw) share one source of truth via
symlinks. Sync `docs/proxy/proxy-load-balancer.md` and
`memory/srs-codebase-map.md` with the new names.

No protocol, log, HTTP API, or wire-format changes. Refactor only — all
externally observable proxy behavior is unchanged.

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-authored-by: chatgpt-codex-connector[bot] <199175422+chatgpt-codex-connector[bot]@users.noreply.github.com>

2026-05-17 12:09:07 -04:00

3 Commits