On Fri, Jun 14, 2019 at 3:02 PM Strahil <hunter86_bg@yahoo.com> wrote:


On Jun 13, 2019 16:14, Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:
>
> Hello,
> I would like to know in better detail how soft fencing works in 4.3.
> In particular, with "soft fencing" we "only" mean vdsmd restart attempt, correct?


Yes, it just restarts vdsmd service using SSH connection. In the past we had several cases, where VDSM was non-responsive, but VMs were running fine, that's why we added this as the 1st step in non-responding treatment flow.
We try to connect to host using SSH, restarts VDSM and waits if host start communicate again. If there is an error during SSH connection or service restart, we immediately continue to next phase of the treatment.

> Who is responsible for issuing the command? Manager or host itself?

The manager should take the decision, but the actual command should be done by another  host.


The manager, this flow is started  from host monitoring if there a network error or connection timeout ...

> Because in case of Manager, if the host has already lost connection, how could the manager be able to do it?

Soft fencing is ussed when ssh is available. In all other cases it doesn't work.


So if engine cannot communicate with host, we don't know the reason, so there are several steps in non-responding treatment:

1. SSH Soft Fencing
2. Kdump detection (if it's configured for the host and we detecte host is dumping, we can restart HA VMs on different host)
3. Power Management restart
    - according to cluster fencing policy we can skip restarting host if for exampl host is renewing its storage lease or gluster cluster is healing
    - this part is executed on different host in the same cluster/data center

If you want to know more about fencing in oVirt, please take a look at below links:

Host fencing in oVirt - Fixing the unknown and allowing VMs to be highly available

Integrating kdump into oVirt

Automatic fencing in oVirt

Fence-kdump integration in oVirt


And course feel free to ask questions

Martin

> Thanks in advance for clarifications and eventually documentation pointers

oVirt DOCs need a lot of updates, but I never found a way to add or edit a page.

Best Regards,
Strahil Nikolov

_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org
Privacy Statement: https://www.ovirt.org/site/privacy-policy/
oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/
List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/OQIENJDAWQNHORWFLSUYWJKH7SS7E5JE/


--
Martin Perina
Manager, Software Engineering
Red Hat Czech s.r.o.