in_tail: group setting judges the limit by the average number, not the total number

[daipom]

Describe the bug

#3535 (comment)

The limit of group setting should control the total number of lines collected for a group (https://docs.fluentd.org/input/tail#less-than-group-greater-than-section).

However, it judges the limit by the average number, not the total number.

To Reproduce

Run Fluentd by the following command with the config below.
(limit: 10 and rate_limit: 1m)

$ rm -f /test/fluentd/pos/pos; echo A-1 > /test/fluentd/input/A.log; echo B-1 > /test/fluentd/input/B.log; \
bundle exec fluentd -c /test/fluentd/config/group/fluent.conf

After running Fluentd, run the following command on another console to add logs to /test/fluentd/input/A.log.

$ for i in `seq 2 9`; do sleep 1; echo A-$i >> /test/fluentd/input/A.log; done;

After 9 seconds, the contents of these log files to be collected are as follows.

• /test/fluentd/input/A.log

A-1
A-2
A-3
A-4
A-5
A-6
A-7
A-8
A-9

• /test/fluentd/input/B.log

B-1

However, the output of Fluentd is as Your Error Log below.
This shows that the data is limited by the average number, not the total number.

After collecting 5 lines from the /test/fluentd/input/A.log, in_tail stops collecting.

2023-05-23 11:41:08.271353203 +0900 test: {"message":"A-1"}
2023-05-23 11:41:08.271511446 +0900 test: {"message":"B-1"}
# Run the command to add logs
2023-05-23 11:41:12.717511479 +0900 test: {"message":"A-2"} 
2023-05-23 11:41:13.721082681 +0900 test: {"message":"A-3"}
2023-05-23 11:41:14.724453738 +0900 test: {"message":"A-4"}
2023-05-23 11:41:15.727744777 +0900 test: {"message":"A-5"}

After 1 minutes, in_tail restarts collecting.

2023-05-23 11:42:08.272066658 +0900 test: {"message":"A-6"}
2023-05-23 11:42:08.272076384 +0900 test: {"message":"A-7"}
2023-05-23 11:42:08.272079574 +0900 test: {"message":"A-8"}
2023-05-23 11:42:08.272082425 +0900 test: {"message":"A-9"}

Expected behavior

The limit of group setting should control the total number of lines collected for a group (https://docs.fluentd.org/input/tail#less-than-group-greater-than-section).

So all 10 lines in To Reproduce must be collected in one period.

Your Environment

- Fluentd version: 1.16.1
- Operating system: Ubuntu 20.04.6 LTS
- Kernel version: 5.15.0-71-generic

Your Configuration

<source>
  @type tail
  tag test
  path /test/fluentd/input/*.log
  pos_file /test/fluentd/pos/pos
  read_from_head true
  refresh_interval 5s
  <group>
    rate_period 1m
    pattern /^(?<log>.*)$/
    <rule>
      match {"log": "/./"}
      limit 10
    </rule>
  </group>
  <parse>
    @type none
  </parse>
</source>

<match test.**>
  @type stdout
</match>

Your Error Log

2023-05-23 11:41:08 +0900 [info]: #0 starting fluentd worker pid=1334352 ppid=1334332 worker=0
2023-05-23 11:41:08 +0900 [info]: #0 following tail of /test/fluentd/input/A.log
2023-05-23 11:41:08.271353203 +0900 test: {"message":"A-1"}
2023-05-23 11:41:08 +0900 [info]: #0 following tail of /test/fluentd/input/B.log
2023-05-23 11:41:08.271511446 +0900 test: {"message":"B-1"}
2023-05-23 11:41:08 +0900 [info]: #0 fluentd worker is now running worker=0
2023-05-23 11:41:12.717511479 +0900 test: {"message":"A-2"}
2023-05-23 11:41:13.721082681 +0900 test: {"message":"A-3"}
2023-05-23 11:41:14.724453738 +0900 test: {"message":"A-4"}
2023-05-23 11:41:15.727744777 +0900 test: {"message":"A-5"}
2023-05-23 11:42:08.272066658 +0900 test: {"message":"A-6"}
2023-05-23 11:42:08.272076384 +0900 test: {"message":"A-7"}
2023-05-23 11:42:08.272079574 +0900 test: {"message":"A-8"}
2023-05-23 11:42:08.272082425 +0900 test: {"message":"A-9"}

Additional context

No response