Description
In production, the app server eventually (And increasingly) runs out of disk space. See e-mail from Ludovic on potential fix. implement and test on Beta. Make sure to document configuration and create roll back process as part of 1.8.10 deployment:
Hi,
Concerning the issue of Disk Space consumption that seems to happen and that
is freed on JVM restart, we have had a known issue at one customer site with
the following scenario:
- Lucene has to merge 2 big indexes and creates a third file
- The 2 initial files are deleted from disk but the disk space is actually
not freed (the files are phantom files) - JVM restarts frees the disk space
For this we have found the following fix which consists in using a different
index memory management java class for Lucene (there are multiple
implementations).
This can be done by adding this to the JVM command line:
-Dorg.apache.lucene.FSDirectory.class=org.apache.lucene.store.MMapDirectory
This could be the problem only if you actually don't see the files. If you
see files that consume the space, please show a list of the files which
would then probably not be from lucene.
Ludovic
Attachments
Issue Links
- depends on
-
XPLUCENE-30 Problem Keeping References to deleted file handles
-
- Closed
-