Gabor Kaszab created IMPALA-13673:
-------------------------------------
Summary: Optimize JVM memory footprint of Iceberg tables in
catalogd cache
Key: IMPALA-13673
URL: https://issues.apache.org/jira/browse/IMPALA-13673
Project: IMPALA
Issue Type: Improvement
Components: Frontend
Reporter: Gabor Kaszab
Assignee: Gabor Kaszab
https://issues.apache.org/jira/browse/IMPALA-11265 takes care of the same for
catalogd.
In coordinator it seems that for Iceberg tables the file descriptors are cached
twice. Once for the related HDFS table and once for the Iceberg table. As a
solution, for the Iceberg table we should only take care of the Iceberg related
part of the file descriptors, and when doing a fromThrift() conversion take the
other part of the file descriptor from the Hdfs table.
Ongoing PR for this: https://gerrit.cloudera.org/#/c/22216
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
