feature request auto failover

82 views

Skip to first unread message

tschend

unread,

Mar 24, 2011, 4:28:18 PM3/24/11

to ganeti

Hi everyone,

i am very interested in auto failover of instances when a node is down
or crashed.

I found ths thread in the discussion list archive:

http://groups.google.com/group/ganeti/browse_thread/thread/8218d10c75cfb46d/88547c37ce256f2a?lnk=gst&q=auto+failover#88547c37ce256f2a

then i found the new feature of OOB in the ganeti 2.4 design docs

http://docs.ganeti.org/ganeti/current/html/design-oob.html

so using the OOB management like IPMI/DRAC/iLO etc would be a good
stonith method?

When node is detacted not responding switch of off and power on the
machines on the secondary nodes.
After that we can maybe run hbal to auto fix the replication.

Any plans on this?

Regards
Thomas

Iustin Pop

unread,

Mar 25, 2011, 11:41:39 AM3/25/11

to gan...@googlegroups.com

On Thu, Mar 24, 2011 at 01:28:18PM -0700, tschend wrote:
> Hi everyone,
>
> i am very interested in auto failover of instances when a node is down
> or crashed.

We too :)

> I found ths thread in the discussion list archive:
>
> http://groups.google.com/group/ganeti/browse_thread/thread/8218d10c75cfb46d/88547c37ce256f2a?lnk=gst&q=auto+failover#88547c37ce256f2a
>
> then i found the new feature of OOB in the ganeti 2.4 design docs
>
> http://docs.ganeti.org/ganeti/current/html/design-oob.html
>
> so using the OOB management like IPMI/DRAC/iLO etc would be a good
> stonith method?

Yes, indeed!

> When node is detacted not responding switch of off and power on the
> machines on the secondary nodes.
> After that we can maybe run hbal to auto fix the replication.
>
> Any plans on this?

Not until now, but this is a very good point. I filled
http://code.google.com/p/ganeti/issues/detail?id=150 to track this.