[ClusterLabs] Q: validate for VirtualDomain

Ken Gaillot kgaillot at redhat.com
Thu Dec 10 13:32:16 EST 2020


On Thu, 2020-12-10 at 16:02 +0100, Ulrich Windl wrote:
> Hi!
> 
> Configuring a VirtualDomain resource I mistyped the config parameter.
> I wonder: Should validate-all check whether the config file exists
> and is readable?

This won't help you now, but I've proposed an extension to the OCF
resource agent standard so that validate-all can use OCF_CHECK_LEVEL
similar to how monitor does currently.

The problem is that there are two use cases for validate-all:

* A GUI or other front-end can validate that a user's proposed
configuration is self-consistent (i.e. no syntax or usage errors). The
tool may not be running on the same host that the resource itself will
be running on, so it can't validate anything on the local host.

* A user may want to know whether a particular node is capable of
running a resource configuration. In this case, validate-all should do
local host checks as well as consistency checks.

My proposal is that an OCF_CHECK_LEVEL of 0 or unspecified would do the
first use case, and an OCF_CHECK_LEVEL of 10 would do the second. Tools
could provide some friendlier way of specifying that.

FYI the current proposal is at:

https://github.com/ClusterLabs/OCF-spec/pull/21

I was planning to merge it into the 1.1 draft spec, then do another PR
to remove the "draft" designation, and bring up that PR on the list for
wider discussion at that point. But comments are welcome now as well.

> The only exception I could imagine is when the config file is
> provided by another resource that isn't running yet.
> Despite of this I wonder why I have to specify the full path of the
> config file, because when using virsh it knows where the
> corresponding configuration files are.
> 
> primitive prm_xen_test-jeos VirtualDomain \
>         params param config="/etc/libvirt/libxl/test-jeos.xml"
> hypervisor="xen:///system" \
>         op start timeout=120 interval=0 \
>         op stop timeout=180 interval=0 \
>         op monitor interval=600 timeout=90
> 
> Regards,
> Ulrich
> 
> 
> 
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
> 
> ClusterLabs home: https://www.clusterlabs.org/
> 
-- 
Ken Gaillot <kgaillot at redhat.com>



More information about the Users mailing list