zhenglin tao created HAWQ-390:
---------------------------------
Summary: Memory leaks in dispatcher: QD uses 48G memory during
ANALYZE lasting 1 ~ 2 days on 100 nodes cluster
Key: HAWQ-390
URL: https://issues.apache.org/jira/browse/HAWQ-390
Project: Apache HAWQ
Issue Type: Bug
Components: Dispatcher
Reporter: zhenglin tao
Assignee: Lei Chang
On about 100 nodes cluster (SFO), we load data to lineitem and then do ANALYZE
on it. It takes 1 ~ 2 days before we cancel the ANALYZE and the QD uses about
48G memory. The major memory consumers are:
```
TopMemoryContext: 48G
----------------------------------------------------------
DispatchDataContext: 43G
Others: 5G
```
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
