Fix resize-volume on RHEL across all datastores

[Petr Malik]

The issues stems from the operating system auto-remounting the volume after it is attached by the taskmanager which then causes the resize (fsck) operation to fail.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/243279

[OpenStack Infra (hudson-openstack)]

Change abandoned by Petr Malik (<email address hidden>) on branch: master
Review: https://review.openstack.org/243279
Reason: This fix has appeared to introduce a race-condition, investigating...

[Joe Chen (joe-chen)]

We encountered the same issue on Debian Jessie guest.

It seems that systemd will create a "var-lib-mysql.mount" (via systemd-fstab-generator) to handle mount/unmount.
Whenever /dev/vdb is attached, this systemd unit will do auto-mounting.

As a workaround, we "mask" this systemd unit BEFORE volume.unmount(mount_point), and "unmask" it AFTER volume.mount(mount_point), then resize function works perfectly.

[Luigi Toscano (ltoscano)]

But resize should be possible online (at least for ext3, ext4 and xfs). Do you (both :) have more detailed logs? Maybe it's something else...

[Joe Chen (joe-chen)]

resize2fs is ok, but "e2fsck -f -p /dev/vdb" needs that vdb is unmounted

[Manoj Kumar (manojnkumar)]

Petr: Do you plan to address the race condition?