Hey,
you''re searching for heartbeat2 with crm.
Thomas
Am Montag, den 09.03.2009, 23:53 -0700 schrieb Nathan
Eisenberg:> Hello,
>
>
>
> I have a new 10 node XEN cluster I''ve built out using 3.3, and a
> dedicated iSCSI SAN. The XEN configuration files are located on a
> volume on the SAN, which is exported to all Dom0s via NFS. The system
> works great, but there is one major issue, and two minor issues which
> I have to deal with before I take this system into production.
>
>
>
> The major issue:
>
> I have not found a method that I like for preventing a DomU from being
> started on two nodes at the same time and clobbering the data.
I''ve
> been considering writing a script to manage lock files which are
> placed into the configuration store directory; these files could
> contain the hostname of the Dom0 where the DomU is running, which
> would solve my first minor issue.
>
>
>
> First minor issue:
>
> I haven''t yet found a method for figuring out where
DomU''s are, short
> of running lots of xm lists or scripting something silly together.
> This hasn''t been a problem in my 2-4 node clusters, but it
won''t work
> as this cluster scales out to its eventual size (20 nodes). If I
> implement the lock file idea, this problem is solvable.
>
>
>
> Second minor issue:
>
> I recall that there was a way to change the behavior of the XEN daemon
> so that it would migrate DomUs on shutdown, rather than suspend them
> to disk. However, I can''t figure out what that was.
>
>
>
> Third minor issue:
>
> Has anyone developed a mechanism for ensuring that VMs are distributed
> evenly throughout the cluster? IE, if I have 10 Dom0s, and 100 of the
> same DomUs with the same memory size and the same load, the mechanism
> should ensure that I have approximately 10 DomUs per Dom0. If a host
> dies, it''d be nice to have something that figures out how to
> distribute the 10 DomUs throughout the cluster evenly, so that each
> Dom0 has 11-12 DomUs on it.
>
>
>
> A lot of this is probably just scripting, but I suspect this is a road
> others have had to walk, and I''d just as soon not reinvent a wheel
> (especially knowing how bad my scripts usually are :-) )
>
>
>
> I apologize for the length of this post!
>
>
>
> Best Regards
>
> Nathan Eisenberg
>
> Sr. Systems Administrator
>
> Atlas Networks, LLC
>
> support@atlasnetworks.us
>
> http://support.atlasnetworks.us/portal
>
>
>
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users