On Thu, Nov 8, 2018 at 2:41 PM <hunter86_bg(a)yahoo.com> wrote:
Hello Community,
Do you have any idea if SBD fencing mechanism will be implemented in the
nearest future?
It will be nice to be able to resset a host just like corosync/pacemaker
cluster implementation.
We have more powerful and reliable mechanism, using host power management.
When power management is not available or fail, we have similar mechanism
using
sanlock, which like sanlock is using the host watchdog and communicating
with
other hosts via shared storage.
You can add a storage lease to HA VM. If the VM loose access to storage on
a host,
the system will start the VM on another host safely, even if the original
host is not
accessible or lost power.
Leases are also used to recover storage jobs (like copying volumes) when
the host
performing the copy became non-responsive. We can fence the storage job and
start
it safely on another host, without waiting for the original host to become
available again.
Can you explain what are the advantages of SBD fencing?
More info:
-
https://ovirt.org/develop/release-management/features/storage/vm-leases/
-
https://www.youtube.com/watch?v=WLnU_YsHWtU
-
https://ahadas.github.io/slides/high_availability_with_no_split_brains.pdf
-
https://www.youtube.com/watch?v=Gh2eB06IE2A
Nir