SIGSEGV in Opentelemetry ouput plugin on termination

[thchan]

Bug Report

Describe the bug

When Fluentbit is terminated while retries are pending in the opentelemetry output plugin, a segmentation fault may randomly occur.

This bug only seems to happen if the input plugins use filesystem storage.

To Reproduce

• Steps to reproduce the problem:

Run the following config:

---

service:
  log_level: debug
  dns.resolver: LEGACY
  flush: 3
  storage.path: /data1/workspace/storage/
  storage.max_chunks_up:    32
  scheduler.base: 2
  scheduler.cap: 200
  
pipeline:
  inputs:
    - name: dummy
      rate: 100000
      storage.type: filesystem

  outputs:
    - name : opentelemetry
      match: '*'
      host:  httpstat.us
      port:  443
      tls:   On
      tls.verify: On
      compress: gzip
      workers:    0
      logs_body_key: message
      logs_uri: /observability/v1/logs
      metrics_uri: /observability/v1/metrics
      traces_uri: /observability/v1/traces
      net.dns.resolver: LEGACY
      net.max_worker_connections:  32
      retry_limit: 10
      storage.total_limit_size: 50M

---

The 404 errors are expected.

After a few minutes, terminate Fluent Bit with ctrl + c.

Fluentbit terminates with a SIGSEGV.

Note: The issue may require a few tries before the segmentation fault is seen.

Expected behavior

Fluentbit should gracefully shutdown every time

Screenshots

Your Environment

• Version used:
  3.2.0

• Configuration:
  Configuration in reproduction steps.

• Environment name and version (e.g. Kubernetes? What version?):
  N/A

• Server type and version:
  N/A

• Operating System and version:
  Ubuntu 22

• Filters and plugins:
  Opentelemetry output plugin