Hi, If you have enabled managed memory, and since all rocksDB instances share the same block cache within one slot, all flink_taskmanager_job_task_operator_window_contents_rocksdb_block_cache_pinned_usage in the same slot would report the same value.
Best Yun Tang ________________________________ From: Chesnay Schepler <ches...@apache.org> Sent: Wednesday, January 27, 2021 20:59 To: 曾祥才 <xcz200...@qq.com>; User-Flink <user@flink.apache.org> Subject: Re: rocksdb block cache usage I don't quite understand the question; all 3 metrics you listed are the same one? On 1/27/2021 9:23 AM, 曾祥才 wrote: hi, all I've enable state.backend.rocksdb.metrics.block-pinned-usage metric , and the flink_taskmanager_job_task_operator_window_contents_rocksdb_block_cache_pinned_usage metric exposed. I'm confused that the total memory used for block cache pinned is sum of flink_taskmanager_job_task_operator_window_contents_rocksdb_block_cache_pinned_usage or just flink_taskmanager_job_task_operator_window_contents_rocksdb_block_cache_pinned_usage (for block cache usage the metric seems per slot)?
