Fluentd creates chunks bigger than chunk_limit_size

[mikromas]

Hi,
I have configured fluentd with kafka output plugin.
Fluentd's configuration is the below:

<store ignore_error>
<buffer topic>
    @type file
    @log_level debug
    path  /data/fluentdlogs/logskafka
    timekey      1d
    flush_thread_count 4
    **chunk_limit_size** 2MB
    overflow_action drop_oldest_chunk
    flush_mode interval
    flush_interval 5s
    total_limit_size 2GB
    max_send_limit_bytes 5000000
</buffer>
@type kafka2
@log_level debug
get_kafka_client_log true
brokers <ip>:<port>
topic_key topic
default_topic lalala
<format>
    @type json
</format>
use_event_time false
required_acks 1
request_timeout 30s
reconnect_on_error true
reload_on_failure true
reload_connections false
</store>

At kafka side I have set message.max.bytes=2000000

I get the below error at fluentd's logs:
2023-09-28 07:42:22 +0000 [warn]: #1 failed to flush the buffer. retry_times=1 next_retry_time=2023-09-28 07:42:24 +0000 chunk="60666729618260cf29737b0db4803708" error_class= Kafka::MessageSizeTooLarge error="Kafka::MessageSizeTooLarge"
but

[td-agent@ztsl-bssc-fluentd-statefulset-2 /]$ ls -lRt /data/fluentdlogs/logskafka/ | grep 60666729618260cf29737b0db4803708
-rw-r--r--. 1 td-agent 2006 2043098 Sep 28 07:42 buffer.q60666729618260cf29737b0db4803708.log
-rw-r--r--. 1 td-agent 2006 84 Sep 28 07:42 buffer.q60666729618260cf29737b0db4803708.log.meta

Seems that the chunk is biger than 2 MB. Why is this since chunk_limit_size is 2MB?

[mikromas]

Any update?