The issue of disk space not being freed has occurred again

[funky-eyes]

What happened?

Unlike last time, after I upgraded the snapshot to the latest 1.0.0 version the issue reappeared. Even when I forced GC with jcmd, disk space was not freed — only restarting the broker released it.

What did you expect to happen?

What else do we need to know?

#513

[AnatolyPopov]

Hi @funky-eyes! Sorry, I did not have time yet to look into this, did you manage to figure out the reason? Should we still look into this?

[funky-eyes]

Hi @funky-eyes! Sorry, I did not have time yet to look into this, did you manage to figure out the reason? Should we still look into this?

I found that the remoteFetch thread is holding the file, preventing the disk space from being freed.

However, I didn't encounter this problem with the earlier 0.0.1-SNAPSHOT build, which used Kafka 3.8.1. The current issue appears in version 1.0.0 together with Kafka 3.9.1.

[AnatolyPopov]

Thanks! Interesting, I will try to find some time to look deeper into this.

[funky-eyes]

Thanks! Interesting, I will try to find some time to look deeper into this.

Thank you. If there's anything else I can assist with, please let me know.

[jiafu1115]

@funky-eyes is it the same issue: apache/kafka#20007. you can double check. if it is same issue. the workarond is to extend the configure item for timeout before new kafka released. cc @AnatolyPopov

[funky-eyes]

@funky-eyes is it the same issue: apache/kafka#20007. you can double check. if it is same issue. the workarond is to extend the configure item for timeout before new kafka released. cc @AnatolyPopov

I don’t think this is the same issue, because what I observed is that disk space is released slowly, but it does eventually get freed. The problem isn’t that the storage space is never released.