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

Posted to notifications@iotdb.apache.org by "Yuan Tian (Jira)" <ji...@apache.org> on 2022/11/08 12:14:00 UTC

[jira] [Created] (IOTDB-4880) Memory leak in query processing

Yuan Tian created IOTDB-4880:
--------------------------------

             Summary: Memory leak in query processing
                 Key: IOTDB-4880
                 URL: https://issues.apache.org/jira/browse/IOTDB-4880
             Project: Apache IoTDB
          Issue Type: Bug
          Components: Core/Query
            Reporter: Yuan Tian
            Assignee: Yuan Tian
         Attachments: image-2022-11-08-20-13-21-385.png

In read and write mixed load, after about 90 mins, the query will be much slower than before and the memory keeps highing and gc will keep active.

After dumping the memory heap, I found that instanceExecution Map in FragmentInstanceManager is too large. Previously, it can be removed by cancel rpc call from Coordinator, but after optimizing, Coordinator won't call cancel rpc if the FI is already in done state. Unfortunately, when FI's state changing to done, it doesn't remove itself from instanceExecution Map in FragmentInstanceManager which causes this memory leak

!image-2022-11-08-20-13-21-385.png!



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