[Linux-ha-dev] Re: [Fwd: Re: Fwd: Questions about CIB]

[Alan Robertson]

Kai Zhao wrote:
> 
> Hi alan ,
> 
> Thank you for your answer, it is what i need. and , now i also have some 
> question about CRM sub-system ;-)
> I have done some test , but i can not find the answer.
> 
> In fact , There are tow condition about the test.
> 
> 
> 1. Resource be stopped by lrmadmin (lrmadmin -E rsc1 stop ....)
> 2. Resource be delete by lrmadmin (lrmadmin -D rsc1)
> 
> Both thease situation i ought to be consider. so there are questions.
> 

Yes.  Right now, we're trying MUCH MUCH more basic tests than these.
> 
> 
> 1. When a resource is stopped by "lrmadmin -E" , whether the resource 
> will be restart by CRM automatic?
>    or The resource will be transform to other node?
> 
>    Now , in fact ,  the resource will not be both recover  and transform 
>  in my test on newest heartbeat 2.0.

OK.  You mean it won't be either recovered on the current node or moved to 
another node?

I would guess that the CIB would have to be configured in such a way that 
it is monitoring the resource.  You probably have not done that.  Since I 
don't know how to configure the CIB to cause the resource to be monitored, 
I can't tell you how to do that :-).

> 
> 2. When i delete a resource using "lrmadmin -D" , what will CRM to do ? 
> Will The resource be re-add by CRM?

This is not a reasonable test, as far as I know.  This is not a failure 
mode which would occur naturally.  And, deleting a resource from the LRM 
just tells the LRM to forget about it.  It doesn't say to stop it.  So, 
doing nothing may be a perfectly correct response.

>    Now , in my test , the resource still running and exist at 
> /var/lib/hearbeat/crm/cib.xml.

LRMadmin is NOT how to delete a resource.  The LRM is not in charge of 
resource policy - the CRM is.  For the most part, lrmadmin is not a very 
interesting or reasonable test tool for testing the whole system.

Stopping a resource with lrmadmin is probably the only reasonable way to 
use lrmadmin to test the whole system, but not the best way to test 
monitoring.  There are far more interesting and subtle failure modes that 
are much more realistic than this.  I believe I said this in my last email.

>    crmadmin -W rsc1 can still find the resource(lrmadmin -L not) .And 
> when you restart heartbeat , the resource will running again.

This is exactly the expected behavior.  The LRM is NOT in charge of 
resources.  Only the CRM should be doing the things that you're doing. 
This is not how to do that test.

I'm not sure how to do that test, but I'm sure Andrew can tell you.

I believe you're going the wrong way when trying to test using lrmadmin. 
It is not a reasonable test tool - except for testing the LRM alone.

By the way, these kinds of discussions should be going on in the -dev 
mailing list, so I've CCed it.

-- 
     Alan Robertson <alanr at unix.sh>

"Openness is the foundation and preservative of friendship...  Let me claim 
from you at all times your undisguised opinions." - William Wilberforce