Description of problem: ----------------------- dm-thinpool disables discards if configured to use with VDO. Its because of this reason, user can't do reclaim space, if VDO volume is used along with thinpool. Version-Release number of selected component (if applicable): ------------------------------------------------------------- RHHI 2.0 How reproducible: ----------------- Always Steps to Reproduce: ------------------- 1. Create a VDO volume 2. Create a thinpool on top of VDO volume 3. Create thin LVs, format it with XFS bricks and use them as gluster bricks 4. Create gluster volume and use it to store VM Images ( using RHV ) 5. Try to perform discard on the fuse mounted filesystem Actual results: --------------- Discard will never reclaim the discarded space Expected results: ----------------- Able to reclaim the discard storage space, to improve storage utilization Additional info: ------------------ There is a RHEL bug - BZ 1600156 - tracking the original issue, this bug is a RHHI side of bug
Removing the devel ack as this bug is not targeted for RHHI 2.0
If this bug is fixed, should we remove the restriction on creating vdo volume with dm-thinp
The dependent RHEL bug is on Modified state. We need to test with RHEL bug fix then will remove otherwise the complete flow will break.
(In reply to Gobinda Das from comment #3) > The dependent RHEL bug is on Modified state. We need to test with RHEL bug > fix then will remove otherwise the complete flow will break. The bug is closed wontfix - with a workaround. So what are the next steps on RHHI?
I think the issue will be fixed in RHEL-8.0 and will have to test with RHEL-8.0 before removing restriction for vdo volume from RHHI deployment side.
As I understand, Configuring LVCache on RHHI is only possible with thin LV's ; Thin LV also are necessary to enable volume snapshots and are currently not possible on VDO volumes (force-created on thick LV by ovirt deployment setup). Does the above comment mean that it will not be possible to configure thin LV for VDO volumes with ovirt-4.3/Centos-7.6 either? We have been postponing rolling out oVirt-4.2 in production because of these two major cons. Please advise if you think RHEL-8 will be necessary in order to have VDO volumes on par with normal volumes (snapshots and LVCache). In that case we will deploy production asap without VDO.
*** Bug 1690608 has been marked as a duplicate of this bug. ***
(In reply to Gobinda Das from comment #5) > I think the issue will be fixed in RHEL-8.0 and will have to test with > RHEL-8.0 before removing restriction for vdo volume from RHHI deployment > side. RHEL 8 support in RHV is going to be very late. To support thinp devices on top of VDO devices, there exists a simple workaround to update the VDO max_discard_sectors. This can be added as part of gluster-ansible and this can fix the issue
The fix is not yet posted and updating the bug status
Updated VDO systemd unit file as per https://access.redhat.com/solutions/3562021 [Unit] Description=VDO volume services After=systemd-remount-fs.service [Service] Type=oneshot RemainAfterExit=yes ExecStartPre=/sbin/modprobe -a uds ExecStartPre=/sbin/modprobe -a kvdo ExecStartPre=/bin/sh -c "echo 4096 > /sys/kvdo/max_discard_sectors" ExecStart=/usr/bin/vdo start --all --confFile /etc/vdoconf.yml ExecStop=/usr/bin/vdo stop --all --confFile /etc/vdoconf.yml [Install] WantedBy=multi-user.target
Addressed via Bug 1693653 *** This bug has been marked as a duplicate of bug 1693653 ***