[
https://issues.apache.org/jira/browse/KAFKA-1327?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13956822#comment-13956822
]
Jay Kreps commented on KAFKA-1327:
----------------------------------
I uploaded a patch that adds the above metrics. Specifically I added
kafka.utils.Throttler.cleaner-io
This is the current-value seen by the I/O throttler on cleaner I/O. It combines
reads and writes across all threads and should be comparable to the I/O
throttle that is configured at the broker level.
kafka.log.Log.topic-partition-Size
This is the total bytes in this partition.
kafka.log.LogCleanerManager.max-dirty-percent
This is the maximum dirty_ratio*100 across all logs.
The following statistics are computed by analyzing the last completed cleaning
for each thread:
kafka.log.LogCleaner.max-buffer-utilization
This is the maximum buffer utilization across all threads in the last cleaning.
kafka.log.LogCleaner.max-clean-time-secs
The maximum time taken to complete a cleaning for each of the threads in their
last cleaning.
kafka.log.LogCleaner.last-cleaning-recopy-rate
The ratio of bytes read to bytes written for the last completed cleaning for
each thread (as a percentage)
> Add log cleaner metrics
> -----------------------
>
> Key: KAFKA-1327
> URL: https://issues.apache.org/jira/browse/KAFKA-1327
> Project: Kafka
> Issue Type: Improvement
> Reporter: Jay Kreps
> Assignee: Jay Kreps
> Attachments: KAFKA-1327.patch
>
>
> The proposal would be to add the following metrics:
> 1. Compaction I/O rate (MB/sec) across all threads (read+write)
> 2. Data size on disk per partition
> 3. Max dirty ratio across all logs
> 4. Max buffer use from all compactions so far
> 5. Survivorship ratio (# records recopied / total records)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
