"na1" should not be used as the nodeagent name. It should be the
hostname of the machine -- and it should be DNS resolvable.
Currently the config for an instance has no special field for the
hostname. DAS needs this to contact the instance. I overloaded the
unused "node-agent-ref" to mean "this is my hostname"
On 5/17/2010 12:29 PM, Bill Shannon wrote:
> As I'm sure you remember, "synchronization" happens in two different
> ways:
>
> - startup synchronization, that synchronizes the instance with the DAS
> when the instance is started
> - command replication, that synchronizes the running instance with the
> DAS by executing the same command on the instance that was executed on
> the DAS
>
> I've got a first cut at startup synchronization ready for more people
> to try.
> For now, it's a separate command. Once I'm sure that it seems to be
> working
> for people, I'll integrate it with the start-local-instance command.
>
> To use startup synchronization, you need some minimal configuration
> information
> on the client. You can either use the create-local-instance command
> if/when
> it's working, or you can create the configuration information yourself
> manually.
> At a minimum, you need this information if you're creating it manually:
>
> mkdir -p nodeagents/na1/agent/config
> echo agent.das.host={das-host} >
> nodeagents/na1/agent/config/das.properties
> mkdir -p nodeagents/na1/in1
>
> "na1" and "in1" are the node agent name and instance name.
> They can be anything. {das-host} is the hostname of the
> machine that will run the DAS (if not localhost).
>
> The nodeagents directory can be anywhere. Normally it will be in the top
> level GlassFish directory next to the domains directory. If it's
> somewhere else, you can use the --agentdir option to specify it.
>
> With this in place, you can do the startup synchronization using:
>
> asadmin _synchronize-instance
>
> The optional operand is the instance name (if there's more than one).
>
> Of course, you can get more details of what it's doing during the sync
> by using AS_DEBUG=true.
>
> Right now it synchronizes only the applications deployed to the instance.
> Since Hong just got --target working for the deploy command, that should
> do what's expected. For testing, you can tell it to synchronize all
> applications to the instance, even if they're not needed, by using the
> --syncallapps option.
>
> Also, application synchronization uses the expanded application directory
> from the DAS. For performance testing there's a --syncarchive option
> that
> will cause it to use the original archive instead. The resulting
> application
> *might* work for a web app, but I definitely do not expect it to work for
> a Java EE app (ear file).
>
> Please try it out and let me know what does and doesn't work for you.
>
> I'm updating my wiki page to reflect the current implementation. You
> can find details of the synchronization criteria here:
> http://wiki.glassfish.java.net/Wiki.jsp?page=V3.1Clustering#section-V3.1Clustering-SynchronizationCriteria
>
>
> As always, feedback is appreciated.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: admin-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: admin-help_at_glassfish.dev.java.net
>
--
Byron Nevins - Oracle Corporation
Home: 650-359-1290
Cell: 650-784-4123
Sierra: 209-295-2188