kill leaves orphaned children

[awronski]

Version: 3.3.1, Ubuntu 18.04

Configuration:

[program:xapi]
command=/usr/bin/npm run start
directory=/opt/nodeapps/xapi
user=nodejs
autostart=true
autorestart=true
startsecs=25
startretries=3
stopasgroup=true
killasgroup=true
stopwaitsecs=10

Process tree after start:

8307 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.21 │  ├─ npm
8317 nodejs     20   0  4632   928   856 S  0.0  0.0  0:00.00 │  │  ├─ sh -c node -r dotenv/config dist/index.js
8318 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.16 │  │  │  └─ node -r dotenv/config dist/index.js
8323 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  │  │     ├─ node -r dotenv/config dist/index.js
8322 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  │  │     ├─ node -r dotenv/config dist/index.js
8321 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  │  │     ├─ node -r dotenv/config dist/index.js
8320 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  │  │     ├─ node -r dotenv/config dist/index.js
8319 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  │  │     └─ node -r dotenv/config dist/index.js
8316 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8315 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8314 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8313 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8312 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8311 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8310 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8309 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  ├─ npm
8308 nodejs     20   0 1098M 41928 25772 S  0.0  0.0  0:00.00 │  │  └─ npm

# supervisorctl stop xapi
Logs:

2018-05-16 11:12:17,518 INFO waiting for xapi to stop
2018-05-16 11:12:19,518 INFO waiting for xapi to stop
2018-05-16 11:12:21,518 INFO waiting for xapi to stop
2018-05-16 11:12:23,518 INFO waiting for xapi to stop
2018-05-16 11:12:25,518 INFO waiting for xapi to stop
2018-05-16 11:12:27,517 WARN killing 'xapi' (21058) with SIGKILL
2018-05-16 11:12:27,518 INFO waiting for xapi to stop
2018-05-16 11:12:27,520 INFO stopped: xapi (terminated by SIGKILL)

Process tree after stop:

8318 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.16 ├─ node -r dotenv/config dist/index.js
8323 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  ├─ node -r dotenv/config dist/index.js
8322 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  ├─ node -r dotenv/config dist/index.js
8321 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  ├─ node -r dotenv/config dist/index.js
8320 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  ├─ node -r dotenv/config dist/index.js
8319 nodejs     20   0  866M 38548 24768 S  0.0  0.0  0:00.00 │  └─ node -r dotenv/config dist/index.js

Thanks in advance,

[awronski]

the app is closing correct when running with systemd

[sam-falvo]

This is a problem I'm encountering as well. For now, I follow up my calls to stop a daemon with a killall command, but it sure would be nice if I didn't have to do that.

(In my case, I'm trying to control a Python Tornado app.)

[the-c0d3r]

I used to have the same problem if my process spawns other processes. It appears that by default supervisor won't try to kill the program's child process. However, after adding killasgroup=true and stopasgroup=true, it doesn't seem to happen anymore.

[Drugoy]

I used to have the same problem if my process spawns other processes. It appears that by default supervisor won't try to kill the program's child process. However, after adding killasgroup=true and stopasgroup=true, it doesn't seem to happen anymore.

But OP provided configuration and there is already stopasgroup=true and killasgroup=true and he still experiences the issue (as well as I do).

[Drugoy]

Any news?

[awronski]

I migrated to systemd

[moooofly]

Any news about this "Grand Children" issue??

[moooofly]

IMO, the key point is "SIGKILL"

Even if with stopasgroup=true, killasgroup=true, stopwaitsecs=10, we can still encounter 'SIGKILL' issue, e.g., target instance can not gracefully quit within 10s, and will be killed after that.

When SIGKILL do its job, something wired will happen. One of them is the result as provided above.

[dat-h-ammann]

I implemented a possible solution which works well for my use-cases. Any reviews, suggestions or comments are appreciated.