[jira] [Comment Edited] (KAFKA-12555) Log reason for rolling a segment

["Alexander Daskalov (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-12555?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17318172#comment-17318172 ] 

Alexander Daskalov edited comment on KAFKA-12555 at 4/9/21, 5:56 PM:
---------------------------------------------------------------------

Actually, looking closer at the current debug line:
{code:java}
debug(s"Rolling new log segment (log_size = ${segment.size}/${config.segmentSize}}, " +
 s"offset_index_size = ${segment.offsetIndex.entries}/${segment.offsetIndex.maxEntries}, " +
 s"time_index_size = ${segment.timeIndex.entries}/${segment.timeIndex.maxEntries}, " +
 s"inactive_time_ms = ${segment.timeWaitedForRoll(now, maxTimestampInMessages)}/${config.segmentMs - segment.rollJitterMs})."){code}

 It is already debugging the reason to roll.
 Each entry in the log (log_size, offset_index_size, time_index_size,inactive_time_ms) print the current and the max value, so you can see what was the reason by checking which current value is above the max value. 
 For example in the line

 
{code:java}
Rolling new log segment (log_size = 90/140}, offset_index_size = 0/1310720, time_index_size = 1/873813, inactive_time_ms = 12357342468/100{code}
you can tell the reason is because of exceeding time. 

However, there is 1 rolling condition which is missing and that's when offsets can't be converted to relative offsets.
 So i'd suggest adding to the log
{code:java}
s"offset_convertible_to_relative = ${segment.canConvertToRelativeOffset(appendInfo.lastOffset)})."){code}
in order to have all 5 reasons in the existing log 

 


was (Author: comonut):
Actually, looking closer at the current debug line:
 debug(s"Rolling new log segment (log_size = ${segment.size}/${config.segmentSize}}, " +
        s"offset_index_size = ${segment.offsetIndex.entries}/${segment.offsetIndex.maxEntries}, " +
        s"time_index_size = ${segment.timeIndex.entries}/${segment.timeIndex.maxEntries}, " +
        s"inactive_time_ms = ${segment.timeWaitedForRoll(now, maxTimestampInMessages)}/${config.segmentMs - segment.rollJitterMs}).")
It is already debugging the reason to roll.
Each entry in the log (log_size, offset_index_size, time_index_size,inactive_time_ms) print the current and the max value, so you can see what was the reason by checking which current value is above the max value. 
For example in the line

 
{code:java}
Rolling new log segment (log_size = 90/140}, offset_index_size = 0/1310720, time_index_size = 1/873813, inactive_time_ms = 12357342468/100{code}
you can tell the reason is because of exceeding time. 

However, there is 1 rolling condition which is missing and that's when offsets can't be converted to relative offsets.
So i'd suggest adding to the log


{code:java}
s"offset_convertible_to_relative = ${segment.canConvertToRelativeOffset(appendInfo.lastOffset)})."){code}
in order to have all 5 reasons in the existing log 

 

> Log reason for rolling a segment
> --------------------------------
>
>                 Key: KAFKA-12555
>                 URL: https://issues.apache.org/jira/browse/KAFKA-12555
>             Project: Kafka
>          Issue Type: Improvement
>            Reporter: Stanislav Kozlovski
>            Assignee: Alexander Daskalov
>            Priority: Minor
>              Labels: newbie
>
> It would be useful for issue-diagnostic purposes to log the reason for why a log segment was rolled (https://github.com/apache/kafka/blob/e840b03a026ddb9a67a15a164d877545130d6e17/core/src/main/scala/kafka/log/Log.scala#L2069)



--
This message was sent by Atlassian Jira
(v8.3.4#803005)