排查SRS问题记录一（RTMP推流，RTC播放卡住）

问题描述：

SRS Version=4.0.126. 配置文件配置：

listen              1935;
max_connections     100;
daemon              off;
srs_log_tank        console;http_server {enabled         on;listen          8080;dir             ./objs/nginx/html;
}http_api {enabled         on;listen          1985;
}
stats {network         0;
}
rtc_server {enabled on;# Listen at udp://8000listen 8000;## The $CANDIDATE means fetch from env, if not configed, use * as default.## The * means retrieving server IP automatically, from all network interfaces,# @see https://github.com/ossrs/srs/wiki/v4_CN_RTCWiki#config-candidatecandidate 192.168.71.35; #本机IP
}vhost __defaultVhost__ {rtc {enabled     on;}http_remux {enabled     on;mount       [vhost]/[app]/[stream].flv;}http_hooks {enabled on;on_rtc_play http://127.0.0.1/control/relay/subscriber;}
}

注释：on_rtc_play http://127.0.0.1/control/relay/subscriber; 自定义开发，目的是在rtc播放拉流时，如果srs服务中没有直播流，可以通过回调到Nginx-RTMP触发RTMP推流到SRS中。

一、使用ffmpeg推流：

ffmpeg -re -i ~/Downloads/test.mp4 -c copy -f flv rtmp://127.0.0.1/live/livestream

二、使用http://127.0.0.1:8080/中SRS播放器播放。播放正常。

三、再打开一个页面使用http://127.0.0.1:8080/中SRS播放器播放，发现无法播放，并且第二部中的播放页面也会卡住。

只要开启rtc enabled on;即使不使用rtc播放，也存在如下问题

使用ffmpeg推流
使用ffplay播放流
使用ffmpeg推相同的流
第2步中的流可以播放，当时关闭播放再使用ffplay播放会失败。

问题分析

一、查看日志，发现有：

cleanup when unpublish
cleanup when unpublish, created=%u, deliver=%u

说明有触发推流断开。猜想可能是第二次进行RTC播放时，回调了http://127.0.0.1/control/relay/subscriber，使Nginx-RTMP服务又推流到SRS，并且app和name是相同的。这样第二次的RTMP推流由于是重复的流，所以会推流失败，然后进入unpublish调用链中。

二、在void SrsLiveSource::on_unpublish()函数中，将RTMP转RTC的bridger_进行了释放。

void SrsLiveSource::on_unpublish()
{。。。。。略if (bridger_) {bridger_->on_unpublish();srs_freep(bridger_);}。。。。。略
}

三、对于RTMP推流，是会经常出现重复推流的情况的。在Nginx-RTMP服务中，只会将第二次重复推流进行禁止，并不会影响第一次已经推上来的直播流。同理，SRS中的RTMP服务应该也有类似的逻辑。于是查看代码：

srs_error_t SrsRtmpConn::publishing(SrsLiveSource* source)
{srs_error_t err = srs_success;SrsRequest* req = info->req;if (_srs_config->get_refer_enabled(req->vhost)) {if ((err = refer->check(req->pageUrl, _srs_config->get_refer_publish(req->vhost))) != srs_success) {return srs_error_wrap(err, "rtmp: referer check");}}if ((err = http_hooks_on_publish()) != srs_success) {return srs_error_wrap(err, "rtmp: callback on publish");}// TODO: FIXME: Should refine the state of publishing.if ((err = acquire_publish(source)) == srs_success) {// use isolate thread to recv,// @see: https://github.com/ossrs/srs/issues/237SrsPublishRecvThread rtrd(rtmp, req, srs_netfd_fileno(stfd), 0, this, source, _srs_context->get_id());err = do_publishing(source, &rtrd);rtrd.stop();}// whatever the acquire publish, always release publish.// when the acquire error in the midlle-way, the publish state changed,// but failed, so we must cleanup it.// @see https://github.com/ossrs/srs/issues/474// @remark when stream is busy, should never release it.if (srs_error_code(err) != ERROR_SYSTEM_STREAM_BUSY) {release_publish(source);}http_hooks_on_unpublish();return err;
}

可以看到只有srs_error_code(err) != ERROR_SYSTEM_STREAM_BUSY才会将RTMP publish连接进行释放。

四、断点查看生成error code的代码

srs_error_t SrsRtmpConn::acquire_publish(SrsLiveSource* source)
{srs_error_t err = srs_success;SrsRequest* req = info->req;// @see https://github.com/ossrs/srs/issues/2364// Check whether GB28181 stream is busy.
#if defined(SRS_GB28181)if (_srs_gb28181 != NULL) {SrsGb28181RtmpMuxer* gb28181 = _srs_gb28181->fetch_rtmpmuxer(req->stream);if (gb28181 != NULL) {return srs_error_new(ERROR_SYSTEM_STREAM_BUSY, "gb28181 stream %s busy", req->get_stream_url().c_str());}}
#endif// Check whether RTC stream is busy.
#ifdef SRS_RTCSrsRtcSource *rtc = NULL;bool rtc_server_enabled = _srs_config->get_rtc_server_enabled();bool rtc_enabled = _srs_config->get_rtc_enabled(req->vhost);if (rtc_server_enabled && rtc_enabled && !info->edge) {if ((err = _srs_rtc_sources->fetch_or_create(req, &rtc)) != srs_success) {return srs_error_wrap(err, "create source");}if (!rtc->can_publish()) {return srs_error_new(ERROR_RTC_SOURCE_BUSY, "rtc stream %s busy", req->get_stream_url().c_str());}}
#endif// Check whether RTMP stream is busy.if (!source->can_publish(info->edge)) {return srs_error_new(ERROR_SYSTEM_STREAM_BUSY, "rtmp: stream %s is busy", req->get_stream_url().c_str());}// Bridge to RTC streaming.
#if defined(SRS_RTC) && defined(SRS_FFMPEG_FIT)if (rtc) {SrsRtcFromRtmpBridger *bridger = new SrsRtcFromRtmpBridger(rtc);if ((err = bridger->initialize(req)) != srs_success) {srs_freep(bridger);return srs_error_wrap(err, "bridger init");}source->set_bridger(bridger);}
#endif// Start publisher now.if (info->edge) {return source->on_edge_start_publish();} else {return source->on_publish();}
}

发现其在这一步直接返回错误了：

if (!rtc->can_publish()) {return srs_error_new(ERROR_RTC_SOURCE_BUSY, "rtc stream %s busy", req->get_stream_url().c_str());}

这是由于在第一个推流时，已经将RTMP转RTC的bridger_的SrsRtcSource中is_created_ = true;所以会返回RTC的推流已经存在。错误码并不是RTMP重复推流的错误码ERROR_SYSTEM_STREAM_BUSY。

问题修复

将acquire_publish中判断RTC推流重复和RTMP推流重复调换位置。如下：

srs_error_t SrsRtmpConn::acquire_publish(SrsLiveSource* source)
{srs_error_t err = srs_success;SrsRequest* req = info->req;// @see https://github.com/ossrs/srs/issues/2364// Check whether GB28181 stream is busy.
#if defined(SRS_GB28181)if (_srs_gb28181 != NULL) {SrsGb28181RtmpMuxer* gb28181 = _srs_gb28181->fetch_rtmpmuxer(req->stream);if (gb28181 != NULL) {return srs_error_new(ERROR_SYSTEM_STREAM_BUSY, "gb28181 stream %s busy", req->get_stream_url().c_str());}}
#endif// Check whether RTMP stream is busy.if (!source->can_publish(info->edge)) {return srs_error_new(ERROR_SYSTEM_STREAM_BUSY, "rtmp: stream %s is busy", req->get_stream_url().c_str());}// Check whether RTC stream is busy.
#ifdef SRS_RTCSrsRtcSource *rtc = NULL;bool rtc_server_enabled = _srs_config->get_rtc_server_enabled();bool rtc_enabled = _srs_config->get_rtc_enabled(req->vhost);if (rtc_server_enabled && rtc_enabled && !info->edge) {if ((err = _srs_rtc_sources->fetch_or_create(req, &rtc)) != srs_success) {return srs_error_wrap(err, "create source");}if (!rtc->can_publish()) {return srs_error_new(ERROR_RTC_SOURCE_BUSY, "rtc stream %s busy", req->get_stream_url().c_str());}}
#endif// Bridge to RTC streaming.
#if defined(SRS_RTC) && defined(SRS_FFMPEG_FIT)if (rtc) {SrsRtcFromRtmpBridger *bridger = new SrsRtcFromRtmpBridger(rtc);if ((err = bridger->initialize(req)) != srs_success) {srs_freep(bridger);return srs_error_wrap(err, "bridger init");}source->set_bridger(bridger);}
#endif// Start publisher now.if (info->edge) {return source->on_edge_start_publish();} else {return source->on_publish();}
}