Healthcheck creating infinite zombie processes #2311
Comments
|
Your non-default configuration of the container/internal ports is problematic:
Leave those at the following and put back the same in If you're still experiencing problems, provide the output of and also add |
|
Since mc-health is simply a bash script, one more thing to adjust in order to debug is to add the following to your With that, again grab the output of health inspection, which will now include the execution debugs of the script: |
|
I seem to have a similar/same issue. My container lags the whole server out with a bunch of |
|
Interesting, but @fisher60 it looks like you're not using autopause, right? |
|
I am not using autopause. Here is the output of adding that healthcheck line to my docker-compose.yml and using docker inspect {
"Status": "starting",
"FailingStreak": 11,
"Log": [
{
"Start": "2023-08-07T20:20:47.5634729Z",
"End": "2023-08-07T20:21:17.634731043Z",
"ExitCode": -1,
"Output": "Health check exceeded timeout (30s): + . /start-utils\n+ '[' -f /data/.mc-health.env ']'\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ mc-monitor status --host localhost --port 25566\n"
},
{
"Start": "2023-08-07T20:21:22.658375119Z",
"End": "2023-08-07T20:21:52.706556752Z",
"ExitCode": -1,
"Output": "Health check exceeded timeout (30s): + . /start-utils\n+ '[' -f /data/.mc-health.env ']'\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ mc-monitor status --host localhost --port 25566\n"
},
{
"Start": "2023-08-07T20:21:57.729941643Z",
"End": "2023-08-07T20:22:27.789959294Z",
"ExitCode": -1,
"Output": "Health check exceeded timeout (30s): + . /start-utils\n+ '[' -f /data/.mc-health.env ']'\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ mc-monitor status --host localhost --port 25566\n"
},
{
"Start": "2023-08-07T20:22:32.809159068Z",
"End": "2023-08-07T20:23:02.866483212Z",
"ExitCode": -1,
"Output": "Health check exceeded timeout (30s): + . /start-utils\n+ '[' -f /data/.mc-health.env ']'\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ mc-monitor status --host localhost --port 25566\n"
},
{
"Start": "2023-08-07T20:23:07.890101611Z",
"End": "2023-08-07T20:23:37.938705575Z",
"ExitCode": -1,
"Output": "Health check exceeded timeout (30s): + . /start-utils\n+ '[' -f /data/.mc-health.env ']'\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ isTrue ''\n+ case \"${1,,}\" in\n+ return 1\n+ mc-monitor status --host localhost --port 25566\n"
}
]
} |
|
Thanks for those logs. With that I noticed that the ping operation itself had a default timeout of 60 seconds, which is longer than the Docker health check timeout. Can you re-pull the latest image and see if that resolves this for you? |
|
Hello, I pulled the latest image and built it. The healthcheck is failing without the zombie processes spawning. So it appears to be fixed now. Thank you! |
|
I think my issue was somehow related or a heck of a coincidence. I pulled latest image and enabled auto_pause. All good. /shrug |
|
@zincarla it was very likely related. I had originally added that timeout in mc-monitor because the paused process would lead to weird socket connectivity. In any case, glad it's resolved for you also. |
Describe the problem
I am having trouble finding the root cause of this problem, but one of my containers was successfully starting and running, but indicating unhealthy and spawning infinite health checks. A new zombie process was being created almost twice per minute and it appears to be originating from the health checks of a specific docker container.
I am unsure if this is due to a specific plugin configuration or odd state achieved. I am unable to reliably reproduce the issue as I run more than 8 nearly identical minecraft servers with a similar or identical plugin stack and similar or identical settings/configurations.
The zombie processes continue unless the container is manually stopped and I only noticed the issue since I actually ran out of process IDs on the bare-metal server and it crashed my entire web server.
As of writing this, the container will not start at all, which is making debugging difficult (this is likely just a result of broken plugins and not related to the above issue).
Container definition
Container logs
The text was updated successfully, but these errors were encountered: