You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Francois Visconte (Jira)" <ji...@apache.org> on 2024/07/23 08:10:00 UTC

[jira] [Updated] (KAFKA-17184) Remote index cache noisy logging

     [ https://issues.apache.org/jira/browse/KAFKA-17184?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Francois Visconte updated KAFKA-17184:
--------------------------------------
    Priority: Critical  (was: Major)

> Remote index cache noisy logging
> --------------------------------
>
>                 Key: KAFKA-17184
>                 URL: https://issues.apache.org/jira/browse/KAFKA-17184
>             Project: Kafka
>          Issue Type: Bug
>            Reporter: Francois Visconte
>            Priority: Critical
>
> We have a tiered storage cluster where some consumers are constantly lagging behind. 
> On this cluster, we get a ton of error logs and fail fetches with the following symptom: 
> {code:java}
> java.lang.IllegalStateException: This entry is marked for cleanup
> 	at org.apache.kafka.storage.internals.log.RemoteIndexCache$Entry.lookupOffset(RemoteIndexCache.java:569)
> 	at org.apache.kafka.storage.internals.log.RemoteIndexCache.lookupOffset(RemoteIndexCache.java:446)
> 	at kafka.log.remote.RemoteLogManager.lookupPositionForOffset(RemoteLogManager.java:1445)
> 	at kafka.log.remote.RemoteLogManager.read(RemoteLogManager.java:1391)
> 	at kafka.log.remote.RemoteLogReader.call(RemoteLogReader.java:62)
> 	at kafka.log.remote.RemoteLogReader.call(RemoteLogReader.java:31)
> 	at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264)
> 	at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
> 	at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
> 	at java.base/java.lang.Thread.run(Thread.java:840)
> {code}
> I believe this should be handled differently:
> * Log should be warn or info
> * We should reload the index when an offset is requested and the entry is marked for cleanup. 
> We do use the default setting for {{remote.log.index.file.cache.total.size.bytes}} (1GiB).



--
This message was sent by Atlassian Jira
(v8.20.10#820010)