Thanks Tom for the suggestion! The cluster devtests that were failing
are now passing. The behavior is different for set and
_register-instance. See below....
On 12/20/2010 9:56 AM, Tom Mueller wrote:
> Jennifer,
>
> Would it work to change this so that the check makes sure that the
> server's config is the same as its cluster's config?
> i.e.,
>
> domain.getClusterForInstance(serverName).getConfigRef().equals(Server.getConfigRef())
>
>
>
> This has to be true for both a set and a _register-instance.
For _register-instance (create-local-instance --cluster c1 i1) the above
is true.
getClusterForInstance(serverName).getConfigRef() = "c1-config"
Server.getConfigRef() ="c1-config"
For set (set servers.server.i1.config-ref=i2-config), the above is false.
getClusterForInstance(serverName).getConfigRef() = "c1-config"
Server.getConfigRef() ="i2-config"
So this does seem to work.....
// cannot change config ref of a clustered instance
Cluster cluster = domain.getClusterForInstance(serverName);
if (cluster != null) { // cluster is not null during
create-local-instance --cluster c1 i1
if (!cluster.getConfigRef().equals(configRef)) {
// During set when trying to change config-ref of a clustered
instance,
// the value of desired config-ref will be different than the
current config-ref.
// During _register-instance, (create-local-instance --cluster
c1 i1)
// cluster.getConfigRef().equals(configRef) will be true
// and not come here.
logger.warning(localStrings.getLocalString("configref.clusteredinstance",
"Cannot change a config-ref when the instance is part of
a cluster"));
return false;
}
}
>
> BTW, the configref.invalid key being used in the custom validator
> might cause some problems with the new MessageInterpolatorImpl that
> went in for MS7. To fix this, try adding ConfigRefConstraint.class to
> the default value for the payload method.
Thanks I was going to ask you about that. I tried your suggestion, bu
it doesn't let me since ConfigRefConstaint does not extend Payload, and
I couldn't extend Payload because ConfigRefConstraint is an interface.
Do you know if the MessageIntepolatorImpl will be fixed?
Thanks,
Jennifer
> Also, the configref.invalid message string should go in the
> LocalStrings.properties file that is in the
> com.sun.enterprise.config.serverbeans.customvalidators package.
>
> Tom
>
> On 12/19/2010 4:53 PM, Jennifer Chou wrote:
>> I'm trying to create a custom constraint/validator for configRef on
>> the Server config bean.
>> There are 2 cases:
>> 1) cannot change config ref of a clustered instance
>> 2) cannot use a non-existent config
>>
>> which only need to be validated when the server instance already
>> exists in domain.xml, like when using the set command.
>> And should not be validated during _register-instance when the server
>> hasn't been created yet. So the following commands should NOT fail
>> create-local-instance --cluster c1 i1
>> create-local-instance i1
>>
>> but these should fail
>> set servers.server.i2.config-ref=c1-config (if i2 is a clustered
>> instance)
>> set servers.server.i2.config-ref=xxx
>>
>> Anybody know if there's maybe a way I can tell inside the custom
>> validator if the server already exists in domain.xml?
>>
>> Issues:
>> http://java.net/jira/browse/GLASSFISH-15087
>> http://java.net/jira/browse/GLASSFISH-15218
>>
>>
>> Thanks,
>> Jennifer