Dear All, We have been noticing an issue with our long-running Kylin clusters (Kylin 3.1.0 and HBase 1.4.10 / EMR 5.28). As our data grows, the HBase region servers are running out of disk space. This seems to be happening because Kylin is configuring a uniquely named coprocessor jar for each table. HBase region servers are downloading these jars to a tmp folder but probably because each table has a uniquely named jar, this coprocessor is now duplicated as many times as there are tables created by Kylin. This issue has been reported in KYLIN-5022 <https://issues.apache.org/jira/browse/KYLIN-5022> by someone else and I also added some findings.
For now, the only workaround I found was extending the disk size for our clusters but that is not a great solution for scaling and cost perspective. Is there a way to reconfigure this behavior? Is it even intentional to use a unique jar name for every table? Regards, Gabor
