You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@zookeeper.apache.org by "Yan Zhao (Jira)" <ji...@apache.org> on 2024/01/25 05:02:00 UTC

[jira] [Created] (ZOOKEEPER-4794) Reduce the ZKDatabase#committedLog memory usage

Yan Zhao created ZOOKEEPER-4794:
-----------------------------------

             Summary: Reduce the ZKDatabase#committedLog memory usage
                 Key: ZOOKEEPER-4794
                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4794
             Project: ZooKeeper
          Issue Type: Improvement
          Components: server
    Affects Versions: 3.9.1
            Reporter: Yan Zhao


In ZKDatabase, after a quorum request is committed successfully, the ZKDatabase will wrap the request into a proposal and store it in the committedLog. 
The wrap operation: Serialize the request to a byte array and wrap the byte array in the QuorumPacket, so if the request payload size is 1M, the Proposal will occupy 2M memory, which will increase the memory pressure.

The committedLog is used for fast follower synchronization, so we can serialize the request in the synchronization of the processes, no need to serialize the request in advance.

It can reduce half of the memory for committedLog




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