K8S: pid is not 1, automatic detection of docker failed, inotify detection is not automatically enabled.

[jinleileiking]

Description'

Please ensure that the markdown structure is maintained.

1. SRS版本(Version): 4.0release'

Please ensure that the markdown structure is maintained.

1. SRS的日志如下(Log):

Please ensure that the markdown structure is maintained.

No.

Please ensure that the markdown structure is maintained.

1. The configuration of SRS is as follows (Config):

Please ensure that the markdown structure is maintained.

Able to update:

Please ensure that the markdown structure is maintained.

inotify_auto_reload    on;
auto_reload_for_docker on;

Unable to update.

Please ensure that the markdown structure is maintained.

auto_reload_for_docker on;

Replay

Please ensure that the markdown structure is maintained.

1. You can start directly on ACK

Please ensure that the markdown structure is maintained.

Expected behavior (Expect)

Please ensure that the markdown structure is maintained.

srs_detect_docker works well.

Please ensure that the markdown structure is maintained.

Analysis:

Please ensure that the markdown structure is maintained.

I don't know the reason, but it caused the "ps" in the container to not use 1.

[root@srs-origin-0 srs]# ps aux
USER         PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
65535          1  0.0  0.0    956     4 ?        Ss   10:22   0:00 /pause
root           8  0.0  0.0  25296  5120 ?        Ss   10:23   0:00 ./objs/srs -c conf/srs.conf
root          14  0.0  0.0    788   248 ?        Ss   10:23   0:00 tini /usr/sbin/crond -f -L /dev/stdout
root          22  0.0  0.0   1568   448 ?        S    10:23   0:00 /usr/sbin/crond -f -L /dev/stdout
root          49  0.0  0.0  11828  1948 pts/0    Ss   10:26   0:00 /bin/bash
root          97  0.0  0.0  51732  1712 pts/0    R+   10:35   0:00 ps aux

Recently, we have been using containerd instead of Docker. I'm not sure if this is related to that.

That code: err = fr.open("/proc/1/cgroup") seems to be incorrect, right?

The phrase "enable auto reload for docker" is not printed in the log. There might be an issue with this sentence.

However, forcing inotify_auto_reload on; seems to directly detect configmap updates.

Updated log:

{"time":"2021-12-28T10:29:56.536+08:00", "level":"Trace", "pid":8, "cid":"g4944dug", "message":"inotify event wd=1, mask=0x40000100, len=32, name=..2021_12_28_10_29_56.500684997, reload=0"}
{"time":"2021-12-28T10:29:59.536+08:00", "level":"Trace", "pid":8, "cid":"g4944dug", "message":"inotify event wd=1, mask=0x100, len=16, name=..data_tmp, reload=0"}
{"time":"2021-12-28T10:29:59.536+08:00", "level":"Trace", "pid":8, "cid":"g4944dug", "message":"inotify event wd=1, mask=0x80, len=16, name=..data, reload=1"}
{"time":"2021-12-28T10:29:59.536+08:00", "level":"Trace", "pid":8, "cid":"g4944dug", "message":"reload config, signo=1"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"config parse complete"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"srs checking config..."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Warn", "pid":8, "cid":"2v02k593","errno":11,  "message":"stats network use index=0, ip=10.33.0.34, ifname=eth0"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Warn", "pid":8, "cid":"2v02k593","errno":11,  "message":"stats disk not configed, disk iops disabled."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"you can check log by: tail -n 30 -f /var/log/srs/srs.log"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"please check SRS by: ./etc/init.d/srs status"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"reload max_connections success."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"reload http_api success, nothing changed."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"reload http stream success, nothing changed."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"reload rtc server success, nothing changed."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"vhost __defaultVhost__ maybe modified, reload its detail."}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"ingest nothing changed for vhost=__defaultVhost__"}
{"time":"2021-12-28T10:29:59.934+08:00", "level":"Trace", "pid":8, "cid":"2v02k593", "message":"reload config success."}

#1635

TRANS_BY_GPT3

[winlinvip]

When automatic detection is not working, it is possible to manually specify it as "on". Please refer to the comments for details.

# Whether auto reload by watching the config file by inotify.
# default: off
inotify_auto_reload off;

# Whether enable inotify_auto_reload for docker.
# If on, it will set inotify_auto_reload to on in docker, even it's off.
# default: on
auto_reload_for_docker on;

As for the issue of PID not being 1, it needs to be investigated. It is possible that some automatic detection functions for Docker may not work properly, making it less convenient to use.

Since it is a non-blocking issue and not very common, if anyone encounters it, please feel free to submit a patch.

TRANS_BY_GPT3

[winlinvip]

Found that Tencent Cloud's K8s (TKE) is not using PID 1. Reference:

https://github.com/ossrs/srs-tke-template

Added a new configuration item that can force the use of Docker, regardless of whether it is successfully detected or not.

# Whether in docker. When SRS starting, it will detect the docker, however
# it might detect failed, then read this config.
# Default: off
in_docker on;

This way, at least when the detection fails, it can be bypassed.

TRANS_BY_GPT3

[winlinvip]

There is no good way to solve this problem.

TRANS_BY_GPT3