xinyiZzz commented on PR #9350: URL: https://github.com/apache/incubator-doris/pull/9350#issuecomment-1118268486
> > > > > _block_mem_tracker > > > > > > > > > > > > So why remove `counterpart`, `counterpart` is just for `_block_mem_tracker`. > > > > > > > > > It seems that _block_mem_tracker tracks memory for vectorized block? If so, we should track them via ExecNode::_mem_tracker， and it should be tracked via hook, right? btw, we should track mmap allocated memory. > > > > > > mmap has been tracked and introduced in pr: #9145 > > `_block_mem_tracker` only tracks memory for vectorized block, `ExecNode::_mem_tracker` includes other memory in ExecNode in addition to memory for vectorized block. This is the purpose of my previous design, so my reply above is whether it is necessary. > > Yep. I understand what you mean. IMHO, It should be case by case, e.g. for the OlapScanNode, the majority memory should be contributed by blocks, so value of block_mem_tracker is about equal value of ExecNode::_mem_tracker？ But for the AggregateNode, may be the majority memory should be contributed by hash table, so we should track memory usage of hash table. I agree with you that the `_block_mem_tracker` of `scan node` is really unnecessary. After the new MemTracker is used for a period of time, the more detailed tracker and the existing virtual tracker can be modified. A little leak in the code: volap_scan_node.h:67 - remove `_block_mem_tracker` definition. volap_scan_node.cpp:323 - remove `_block_mem_tracker` create. -- 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: commits-unsubscr...@doris.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org For additional commands, e-mail: commits-h...@doris.apache.org
