srowen commented on issue #24461: [SPARK-27434][CORE] Fix mem leak due to
hadoop fs caching mechanism when eventLog is enabled
URL: https://github.com/apache/spark/pull/24461#issuecomment-488030569
Hm, I wonder if that's more at the root of the problem: you'd generally not
close the
srowen commented on issue #24461: [SPARK-27434][CORE] Fix mem leak
URL: https://github.com/apache/spark/pull/24461#issuecomment-487657526
Let's see what tests say so far. BTW you'll need to improve the title of the
PR and JIRA.
srowen commented on issue #24461: [SPARK-27434][CORE] Fix mem leak
URL: https://github.com/apache/spark/pull/24461#issuecomment-487422243
Hm your heap dump is convincing, although this `FileSystem` is just obtained
from Hadoop and its cache. Generally that's not safe to close, but maybe
srowen commented on issue #24461: [SPARK-27434][CORE] Fix mem leak
URL: https://github.com/apache/spark/pull/24461#issuecomment-487390580
I don't think you can close `FileSystem` objects. They're shared. Why is it
a memory leak?