TakaHiR07 opened a new issue, #4915: URL: https://github.com/apache/bookkeeper/issues/4915
**BUG REPORT** I have found that bookie process memory keep improve for a long time, which result in linux kern OOM. The JVM config of our bookies is heap memory 6G,direct memory 10G. <img width="1724" height="361" alt="Image" src="https://github.com/user-attachments/assets/4227fe38-70bf-4f71-a182-5f7bc19e63fc" /> I use `pmap` to analyse the memory statistic of bookie process. And found there are so many glibc arena memory. `pmap` result is as follow: [pmap.txt](https://github.com/user-attachments/files/34169869/pmap.txt) | Mapping | RSS | Meaning | |---|---|---| | 5128 MB anonymous segment | 4969 MB | Nearly fully utilized; direct buffer allocated at startup for write cache and read cache | | 158 anonymous segments (32–64 MB each) | 3214 MB | glibc allocation areas; 38 of these are essentially full (~2.3 GB total), while the remaining 93 have an RSS of less than 10 MB each | This glibc arena is mainly generated by rocksdb. However, our rocksdb config use the default config, it should not generate GB level memory. Further more, it is likely that 64MB block is the memory which is free, but glibc keep in RSS and do not return to linux. And this glibc memory keep improve. A clear and concise description of what the bug is. ***To Reproduce*** Steps to reproduce the behavior: 1. set up a pulsar cluster 2. keep write msg to bookie. make the throughput stable and high. 3. the higher throughput write to the bookie, the faster memory usage of the bookie process grows ***Expected behavior*** With stable throughput, bookie memory usage should also keep stable. ***Screenshots*** ***Additional context*** Add any other context about the problem here. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
