Memory Limit Not Enforced When Using Filesystem Storage

[yuliia-fryshko]

Bug Report

Describe the bug

We are encountering an issue with Fluent Bit when using the filesystem storage. Our intention is to ensure that when Fluent Bit experiences an overload, logs will be stored in a file to prevent data loss. However, during our tests, it appears that Fluent Bit is not respecting the configured memory limits and continues to consume memory despite being configured to offload logs to the filesystem.

To Reproduce

• Rubular link if applicable:
• Example log message if applicable:

{"log":"YOUR LOG MESSAGE HERE","stream":"stdout","time":"2018-06-11T14:37:30.681701731Z"}

• Steps to reproduce the problem:

Here is the configuration of fluentbit we used

[SERVICE]
    Grace                        30
    Flush                        1
    Daemon                       Off
    HTTP_Server                  On
    HTTP_Listen                  0.0.0.0
    HTTP_PORT                    2020
    Parsers_File                 parsers.conf
    storage.path                 /var/log/flb-storage/
    storage.metrics              On
    storage.max_chunks_up        2
    storage.backlog.mem_limit    5M
    storage.buffer_chunk_size    2M
    storage.buffer_max_chunks_up 2
    storage.total_limit_size     50M
    log_level                    trace

[INPUT]
    Name                http
    Listen              0.0.0.0
    Port                24224
    storage.type        filesystem

[METRICS]
    Name            prometheus
    Host            0.0.0.0
    Port            2021
    Scrape_Interval 1

[OUTPUT]
   Name file
   Format template
   Template {time} used={Mem.used} free={Mem.free} total={Mem.total}

1. Run Fluent Bit with the above configuration.
2. Send a high volume of logs using the following script:

import requests
import json
import argparse
from datetime import datetime
import time


def send_log(url, log_data):
 headers = {'Content-Type': 'application/json'}
 response = requests.post(url, headers=headers, data=json.dumps(log_data))
 if response.status_code == 200:
   print(f"Logs sent successfully")


def generate_large_log_message(index, size_in_kb):
 large_log = "A" * (size_in_kb * 1024 - 200)  # Adjust the size for the rest of the JSON structure
 return {
   "container_name": "/ecs-appId-dec89697b7bcbbaaed01",
   "source": "stdout",
   "log": f"This is log number {index} at {datetime.now().isoformat()} {large_log}",
   "container_id": "ef7f185cd674dd5969f8f1c2da2a953af355739329f9137ac"
 }


def send_logs(url, logs_per_second, log_size_kb):
 log_index = 0
 while True:
   start_time = time.time()
   for _ in range(logs_per_second):
     log_data = generate_large_log_message(log_index, log_size_kb)
     send_log(url, log_data)
     log_index += 1
   elapsed_time = time.time() - start_time
   time_to_sleep = max(0, 1 - elapsed_time)
   time.sleep(time_to_sleep)


if __name__ == "__main__":
 parser = argparse.ArgumentParser(description='Send logs to Fluent Bit.')
 parser.add_argument('--logs_per_second', type=int, required=True, help='Number of logs to send per second')
 parser.add_argument('--log_size_kb', type=int, required=True, help='Size of each log in KB')
 parser.add_argument('--url', type=str, default='http://localhost:24224', help='Fluent Bit HTTP input URL')


 args = parser.parse_args()


 send_logs(args.url, args.logs_per_second, args.log_size_kb)

3. Execute the script with the following command:

python3 send_logs.py --logs_per_second 1000 --log_size_kb 1000 --url[ http://localhost:24224](http://localhost:24224/)

Observed behavior
The memory usage of the Fluent Bit container continues to grow despite the configuration of storage.max_chunks_up and storage.backlog.mem_limit.
Files are being written to the filesystem as .flb files in /var/log/flb-storage/, but the logs are also retained in memory, causing the memory to increase.

Expected behavior
Fluent Bit should respect the storage.max_chunks_up and storage.backlog.mem_limit and offload logs to the filesystem once the memory limit is reached, preventing further memory growth.

[github-actions]

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days. Maintainers can add the exempt-stale label.

[github-actions]

This issue was closed because it has been stalled for 5 days with no activity.