Using workspaces for "ALMOST" static data

Question

          Hi,
          The application that we are developing has the following requirement:
          We have a whole bunch of data that is ALMOST static as far as the application is concerned.
          However this data can change infrequently.
          We have two solutions in mind
          1) We are planning to have a StaticDataManager component that will handle the reading/updation of this data.
          All the other components of the application will get the static data from the StaticDataManager
          and cache this data to avoid communication between the components.
          The application will be hosted on a clustered WLS configuration.
          Now suppose an updation takes place, how best can we update all the components that have the cached data?
          2) In case we use Workspaces to cahce this static data, does the cluster configuration take care of
          updating the workspaces in the machines in the cluster?
          TIA!
          Sreeja
          

Answer 1

If you do like you say and specify servera and then specify serverb
          respectively, then yes the workspace doesn't show the same information
          across the cluster. However, if you always specify the cluster, serverc,
          then no matter which side you update or read from the data is the same. So
          yes, if you specifiy a specific server in the cluster, and not the cluster,
          the data will not replicate across the cluster, and you lose concurrency.
          Let me paste in the output to show you what I am seeing which proves the
          data is being replicated with a single write.
          Tester ver. 1.00.00 ----> Test started...
          Connected successfully using http to ethouic97/198.171.100.107:80
          ws.store(DOG, FIDO)
          Connected successfully using http to ethouic97/198.171.100.106:80
          ws2.fetch(DOG)
          DOG = FIDO
          These are some system outs from the test app below with a few extra system
          outs. Notice if the cluster is always your connection point and not an
          individual server, in this case ethouic97. Then the answer to the original
          first question is YES, you can store near static data in workspaces on a
          cluster to have the data replicated across the cluster. However, you must
          always address the cluster and not the individual servers, in this case
          address ethouic97 never server 106 or 107 directly.
          Jonathon Cano
          "Mike Reiche" <[email protected]> wrote in message
          news:[email protected]...
          >
          > Well, it does look like it works, doesn't it?
          >
          > Now try running the first half on one WLS instance and the other half of a
          different WL instance and
          > specify servera in the first half and serverb in the second half. If the
          workspace is replicated it should
          > work.
          >
          > It's broken now. Not really. Regardless of where the t3 connection is
          made, the workspace is
          > in the local WL instance.
          >
          > Mike
          >
          >
          > "Jonathon Cano" <[email protected]> wrote:
          > >Details:
          > >We are using a DNS cluster where serverc actually maps to servers a and
          b.
          > >If you set the workspace to SCOPE_SERVER, and use a t3client connection
          to
          > >the serverc, it can not be servera or serverb directly, then the
          workspace
          > >is updated or read from the data is persisted across both sides of the
          > >cluster. All reads and writes must be done using the cluster name for
          this
          > >to work. I use this to manage fail-over for some Vitria subscribers I
          have
          > >runnig at startup. I have a tester I wrote which works, here is the
          code:
          > >
          > > private void testWorkSpaces() {
          > >
          > > try {
          > >
          > > T3User userInfo = new T3User("system","password");
          > >
          > > T3Client t3 = new T3Client("http://cluster:80",userInfo);
          > > t3.connect();
          > >
          > > WorkspaceDef ws =
          >
          >t3.services.workspace().getWorkspace().getWorkspace("test",WorkspaceDef.OPE
          N
          > >,WorkspaceDef.SCOPE_SERVER);
          > > ws.store("DOG", "FIDO");
          > >
          > > t3.disconnect();
          > >
          > > T3Client t32 = new T3Client("http://cluster:80",userInfo);
          > > t32.connect();
          > >
          > > WorkspaceDef ws2 =
          >
          >t32.services.workspace().getWorkspace().getWorkspace("test",WorkspaceDef.OP
          E
          > >N,WorkspaceDef.SCOPE_SERVER);
          > > String dogName = (String) ws2.fetch("DOG");
          > >
          > >
          > > t32.disconnect();
          > >
          > > System.out.println(dogName);
          > >
          > > } catch (Exception e) {
          > > System.out.println("Test Work Spaces Failed.);
          > > e.printStackTrace();
          > > }
          > >
          > > }
          > >
          > >The first t3client connection echos out that it is connecting to servera.
          > >The second t3client connection echos out that it is connecting to
          serverb.
          > >However, the data being returned is the name FIDO which was written to
          the
          > >workspace when connected to servera. Thus, as long as you are connecting
          to
          > >the cluster, you get replication of data across cluster for workspaces.
          > >
          > >WLS 5.1
          > >
          > >Jonathon Cano
          > >
          > >
          > >"mreiche" <[email protected]> wrote in message
          > >news:[email protected]...
          > >>
          > >> >I am confused.
          > >>
          > >> Yes.
          > >>
          > >> >So why was the original answer to workspaces being used for clustering
          > >near static data NO?
          > >>
          > >> Because that's what the documentation says?
          > >>
          > >> Are you really, really sure it's being propagated? Or perhaps it's just
          > >being updated identically on both instances?
          > >>
          > >> Mike
          > >>
          > >> "Jonathon Cano" <[email protected]> wrote:
          > >> >I am confused. I am storing data in the workspaces right now, and I
          am
          > >> >using this data in a cluster. I can update the data in the cluster
          just
          > >> >fine, and both sides of the cluster are being updated. So why was the
          > >> >original answer to workspaces being used for clustering near static
          data
          > >NO?
          > >> >
          > >> >Jonathon Cano
          > >> >
          > >> >"Tao Zhang" <[email protected]> wrote in message
          > >> >news:[email protected]...
          > >> >>
          > >> >> But the problem is whether you want Exactly-once-per-cluster data or
          > >not.
          > >> >> If not, you can deploy the data object into each server in the
          cluster,
          > >> >but in the data object, it contains the logic of updating itself.
          > >> >> Otherwise, you have to make it a pinned service.
          > >> >>
          > >> >> It's hard to describe here.
          > >> >>
          > >> >> Hope the document can solve your problem.
          > >> >>
          > >> >> http://www.weblogic.com/docs51/classdocs/API_jndi.html#user
          > >> >>
          > >> >> Good Luck.
          > >> >>
          > >> >> Tao Zhang
          > >> >>
          > >> >> "Sreeja" <[email protected]> wrote:
          > >> >> >
          > >> >> >Thanks. Can you please elaborate on the solutions 1 and 2. We are
          > >trying
          > >> >to eliminate the
          > >> >> > usage of 3 - entity beans - since we do not want the components
          using
          > >> >the data to contact
          > >> >> >the entity bean every time they need the static data - as this
          would
          > >> >result in remote calls
          > >> >> >for every usage.
          > >> >> >Sreeja
          > >> >> >
          > >> >> >"Tao Zhang" <[email protected]> wrote:
          > >> >> >>No, the workspace can't be clustered, the data can't be propagated
          to
          > >> >other
          > >> >> >>servers in the cluster.
          > >> >> >>Of course , you can select one server as the data holder and other
          > >> >servers
          > >> >> >>will be t3 client. But it's time consuming to
          > >> >> >>connect to the data holder server.
          > >> >> >>Possible solutions:
          > >> >> >>1. RMI.
          > >> >> >>Deploy the rmi as pinned service which hold the data.
          > >> >> >>2. jndi
          > >> >> >>The data can be propagated by jndi's default behavior.
          > >> >> >>3. database or entity bean.
          > >> >> >>
          > >> >> >>Hope this can help.
          > >> >> >>
          > >> >> >>Tao Zhang
          > >> >> >>
          > >> >> >>Sreeja <[email protected]> wrote in message
          > >> >> >>news:[email protected]...
          > >> >> >>>
          > >> >> >>> Hi,
          > >> >> >>>
          > >> >> >>> The application that we are developing has the following
          > >requirement:
          > >> >> >>> We have a whole bunch of data that is ALMOST static as far as
          the
          > >> >> >>application is concerned.
          > >> >> >>> However this data can change infrequently.
          > >> >> >>> We have two solutions in mind
          > >> >> >>> 1) We are planning to have a StaticDataManager component that
          will
          > >> >handle
          > >> >> >>the reading/updation of this data.
          > >> >> >>> All the other components of the application will get the static
          > >data
          > >> >from
          > >> >> >>the StaticDataManager
          > >> >> >>> and cache this data to avoid communication between the
          components.
          > >> >> >>> The application will be hosted on a clustered WLS configuration.
          > >> >> >>> Now suppose an updation takes place, how best can we update all
          the
          > >> >> >>components that have the cached data?
          > >> >> >>>
          > >> >> >>> 2) In case we use Workspaces to cahce this static data, does the
          > >> >cluster
          > >> >> >>configuration take care of
          > >> >> >>> updating the workspaces in the machines in the cluster?
          > >> >> >>>
          > >> >> >>> TIA!
          > >> >> >>> Sreeja
          > >> >> >>>
          > >> >> >>
          > >> >> >>
          > >> >> >
          > >> >>
          > >> >
          > >> >
          > >>
          > >
          > >
          >