We recently encountered a problem with GlusterFS (7.x) when an application used the flock
syscall on a GlusterFS path. If somehow two flock
syscalls are made at the same time, the lock will never be released. And all future flock
syscalls will wait forever.
This problem doesn't happen only on our infrastructure. There are a lot of bugreports, e.g.:
There are also entries regarding flock
or locking in general in each major release notes: