Description of problem: With glusterFS native driver running Kilo, Gluster volume is restarted as part of providing and revoking 'cert' based access in gluster native driver. This disrupts IO to existing mounts. Granting and revoking access should happen more gracefully without any disruption to existing mounts. Version-Release number of selected component (if applicable): openstack-manila-2015.1.0-2 How reproducible: always Steps to Reproduce: 1) Create a manila share from gluster backend 2) allow 'cert' based access to a nova VM 3) Mount the share on the nova VM and run IOs 4) allow 'cert' based access to another nova VM 5) IO on first nova VM will fail Note: The behavior is same while access is denied as well Actual results: IO on any existing mount fails Expected results: There shouldn't be any disruption to IOs while granting and revoking access Additional info: With this issue. administrator will have to unmount shares on all clients before allowing or denying access to a share using glusterFS native driver.
No fix shall be provided for upstream Kilo and thus it's not fixed in RHEL OSP7 either. It's added to the OSP7 Errata, see https://access.redhat.com/documentation/en/red-hat-enterprise-linux-openstack-platform/version-7/release-notes#idm140465730614816 The bug is handled in upstream Liberty. To track the downstream port of that fix, please see the variant of this bug for OSP8: BZ #1263542