Description of problem: Gluster server 3.6.2 works on top of Centos 6.6 without any issues for several months. There are several small sqlite3 databases on NFS-mounted gluster partition on Centos 6.6 server. Suddenly sqlite started to fail producing "Error: database is locked" message even on a newly created database files. There is no any error messages in gluster logs. I created a statedump and clear all locks related to the file, but it did not resolve the issue. Version-Release number of selected component (if applicable): glusterfs-server-3.6.2-1.el6.x86_64 How reproducible: Have no idea how to reproduce it on a different server. Additional info: In a mounted directory it looks like the following: user@server2:/newops/NFSDBTest $ ls -l total 0 user@server2:/newops/NFSDBTest $ sqlite3 test.db SQLite version 3.7.17 2013-05-20 00:56:22 Enter ".help" for instructions Enter SQL statements terminated with a ";" sqlite> .schema Error: database is locked sqlite>
This bug was incorrectly reported against the Red Hat Gluster Storage product. Instead it should have been reported against the Gluster Community version. Andrey, are you still hitting this problem with current versions of Gluster? In general, sqlite is not very happy with NFS mounts. There are some hints about this in https://sqlite.org/wal.html . Some of the locks that sqlite uses are kept in memory, and not on the filesystem.
No, currently we are not suffering from this bug. I've upgraded gluster server to 3.7.4 at some point and problem gone away, either because of upgrade or because of just a reboot.