Open mirko-lazarevic opened 2 months ago
Does anyone have any insights into what might be going wrong here? I've noticed that Fluent Bit suddenly exits when it encounters the error [input chunk] fail to drop enough chunks in order to place new data
. Is this behavior intentional? Shouldn't Fluent Bit shut down gracefully and release resources properly?
...
if (count != 0) {
flb_error("[input chunk] fail to drop enough chunks in order to place new data");
exit(0);
}
cc: @agup006, @ScarletTanager
Can you see if it is reproducible in the latest version? There have been some bug fixes that may affect this.
Can you step up log_level debug
as well to provide more information?
There should be a stack trace as well.
Ah I notice you are using custom golang plugins, my guess is it might be one of those - make sure you're using latest calyptia-plugin
module to fix up segv in hot reload handling/registration (or ensure you have the correct ABI updates in place).
Also is it reproducible without the golang plugins? I don't think we can investigate issues if they are related to custom plugins so it needs to be reproducible without.
Thanks @patrick-stephens , I’ll attempt to obtain the stack trace first, as I’m currently without any clear information.
Honestly I would step up your dependencies as well first - you can run your existing stuff of course to get stack trace but ultimately I bet it's a dependency problem.
Bug Report
Describe the bug
We are experiencing issues with in our Fluent Bit deployments in Kubernetes cluster, where 1 or 2 pods entering a CrashLoopBackOff state on a cluster consisting of 6 to 8 nodes.
To Reproduce
(SIGSEGV)
Fail to drop enough chunks in order to place new data
Expected behavior
No such file or directory
Fail to drop enough chunks in order to place new data
Screenshots
Your Environment
Additional context