No where can I find something concrete to tell me what kernels were tested, I realize my Kernel is out of date (2.6) and so I may run into similar issues as the poster in the discussion.
Reasonably modern versions of the Linux distros on the support matrix will be safe. Distro maintainers do lots of crazy backporting so its honestly pretty hard to tell what is on what kernel version sometimes.
At a glance the linked issue seems to be actually about GlusterFS. I imagine so long as you stay away from shared filesystems you should be fine. I believe our CI server is publicly viewable so you can look at it, but I know that Elasticsearch gets lots of testing on reasonably modern Debian, Ubuntu, Fedora, RHEL, OEL, Windows, and OSX.
The current kernel depends on your Linux distribution. You can check the kernel version with uname --kernel-release
There were bugs with the journal block device layer (jbd2) and ext3/ext4 in older kernels. You will encounter the issue on spindle disks. If you want to have this fixed on kernel level, you should use Linux kernel >= 2.6.35 or a patched vendor kernel of >= 2.6.32. If that is not possible, or you still see the issue discussed, you can modify filesystem mount settings to make journal management more I/O efficient. The noatime mount option is important.
Thank you both Nick and Jorg, we are currently running CentOS 6.5 (2.6.32-431.29.2.el6.x86_64 kernel) which fits within the matrix you referenced. We are using SSD's for the data drives without any shared file systems so we could be in OK shape. I will for sure monitor our upgrade to 2.1 and before then do some reading,
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.