Fluentbit in constant CrashLoopBackOff because of SIGSEGV

Google Cloud Serverless Applications
,
1

For some weeks now, the fluentbit pods that are managed by GKE, have been in a CrashLoopBackOff cycle, causing logs to be delayed into Logs Explorer.

The fluentbit container is constantly SIGSEGV erroring out. Initially I thought that it might’ve been because of too high log throughput, so I followed the guide here Adjust log throughput  |  Google Kubernetes Engine (GKE)  |  Google Cloud Documentation to enable high throughpot logging on the nodes (all nodepools have been re-created). But the issue is still occuring.

Example crash

[2026/01/06 15:54:28] [engine] caught signal (SIGSEGV)
#0  0x5c0717761e4c      in  flb_output_return() at include/fluent-bit/flb_output.h:1092
#1  0x5c0717761e4c      in  flb_output_return_do() at include/fluent-bit/flb_output.h:1119
#2  0x5c0717768a24      in  cb_stackdriver_flush() at plugins/out_stackdriver/stackdriver.c:3076
#3  0x5c0717c33146      in  co_init() at lib/monkey/deps/flb_libco/amd64.c:117
#4  0xffffffffffffffff  in  ???() at ???:0

Running version v1.34.1-gke.3355002 that has fluentbit v3.2.8

Any idea on how to solve this?

1 Like
2

This appear to have been solved in a recent update, at least 1.34.1-gke.3947000 in Regular channel.

1 Like
3

That means the fixes were included in that release. Did you see any release notes related to this topic?

4

Nothing in the release notes mentioning fluent-bit.

But I noticed that it’s now running fluent-bit v4.0.13 and instead of using the stackdriver output, it’s outputting via http to a sidecar running that is relaying it to stackdriver.

And none of he pods has had any crashes since they started, previously that would crash every couple of minutes.

5

Thank you for the information dailydev-olemartin.

6

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.