[ClusterLabs] Antw: Re: Q: "migrating a clone" with crm shell

Ulrich Windl Ulrich.Windl at rz.uni-regensburg.de
Tue Jun 2 14:54:55 UTC 2015


>>> Ken Gaillot <kgaillot at redhat.com> schrieb am 02.06.2015 um 15:41 in Nachricht
<556DB299.6030907 at redhat.com>:
> On 06/02/2015 03:02 AM, Ulrich Windl wrote:
[...]
> 
> globally-unique="false" means that all clone instances are equivalent,
> and only one instance is ever needed on any one node. So there's no
> logical meaning to migrating such a clone. You can stop it on any node
> any time (e.g. by putting the node in standby, which you'd probably do
> for an upgrade anyway).

OK, how do I stop it on one node; with a location constraint as someone said before? Actually I did that, and it worked.
BTW: For the "miigrate" I'd expect that the cluster finds out that the local primitive can't run elsewhere and thus shuts it down.

> 
> If each clone instance handles a different range of input, so that more
> than one instance can run on a single node, then you want
> globally-unique="true".

Which wouldn't make sense in my case.

[...]

Regards,
Ulrich






More information about the Users mailing list