Performance of Clustered Managed Servers

WLS 10.3.5 on Solaris10
we have the clustered weblogic10.3.5 domain with two managed servers, which a web app deployed.
we use proxy server (an ohs server) as the frontend host.
we tested for the app login, if both managed servers are started, the login takes about 2 mins.
however if only one managed server is started, the login only takes less 1/2 min.
There is no problem to support two managed servers in the ram and cpu on the server
Why?
where should we start to look into the issue?

Hi,
According to the info you sent, there are 71GB of swap area in use and 7GB of real memory free.
This is not conclusive, as we don't know what is in the swap area. It could be part or the entire JVM for your application. Also, we have to take into consideration shared memory.
I believe you are running many applications in that server, and possible multiple environments.
In order to avoid spending too much time investigating other config issues that might not exist, I think you should try to eliminate the swap to see if this is the issue. I believe that this could be the main issue, based on the info I have so far.
Would be possible for you to stop all or some of the applications running in that server? Then you should check the available RAM (make sure that there are at least 10GB available before starting both managed servers) and restart both managed servers which you have your application running.
Regarding the response times for the login, does it always take the same time to login? You mentioned about 30 seconds, this also seems very high to me. Are your discarding the first login after the server restart? Depending on which framework you are using, it might be interpreting some artifacts or compiling the JSPs in the first execution.
Regards,
Fabio Douek

Similar Messages

  • Performance Clustered Managed Servers

    I posted it in the cluster thread, want it be visible here as well.
    Performance of Clustered Managed Servers

    Hi,
    According to the info you sent, there are 71GB of swap area in use and 7GB of real memory free.
    This is not conclusive, as we don't know what is in the swap area. It could be part or the entire JVM for your application. Also, we have to take into consideration shared memory.
    I believe you are running many applications in that server, and possible multiple environments.
    In order to avoid spending too much time investigating other config issues that might not exist, I think you should try to eliminate the swap to see if this is the issue. I believe that this could be the main issue, based on the info I have so far.
    Would be possible for you to stop all or some of the applications running in that server? Then you should check the available RAM (make sure that there are at least 10GB available before starting both managed servers) and restart both managed servers which you have your application running.
    Regarding the response times for the login, does it always take the same time to login? You mentioned about 30 seconds, this also seems very high to me. Are your discarding the first login after the server restart? Depending on which framework you are using, it might be interpreting some artifacts or compiling the JSPs in the first execution.
    Regards,
    Fabio Douek

  • Multiple managed servers, Multiple machines and Multiple clusters

    Hi,
    I have couple of applications to be deployed in high availabliity. I
    have been a Tomcat/Apache guy so far. Yesterday night I read the whole admin doc
    and tried some installations on the boxes. But have some confusion here.
    Since our application is getting commercialised and I like to have 12 managed
    serves on each of the three big linux servers and one windows machine as admin
    server port 9002. 18 for each application in clustered environment. cluster of
    18 for each application.
    FIRST CONFUSION is
    I want to use only one domain name. So I install weblogic81
    on win box and call it dls and then I find out that I need to install weblogic81
    as good as the one on win box on linux servers, not surpised, but during the installation
    when it asked me for a domain name I was confused. No option to join the existing
    domain that I created on my winbox.
    Why??? So I type dls again.
    Is there a way not to do all this as it may be creatign another domain and just
    start nodemanager and not install domain at all ?
    SECOND CONFUSION is
    When I was installing weblogic on win machine as admin server. I was given
    options of adding managed servers and clusters amd machines. So do we need to
    have all the managed servers already running before I add them here or I can do
    it later.
    Thanks n Advance.
    Vivke

    http://www.rittmanmead.com/2010/07/obiee-11gr1-scaleout-clustering-and-high-availability/
    hope it helps

  • Transaction span two WLS managed servers (non-clustering)

    (Weblogic 6.1 - non-clustering version)
    I have two managed servers configured on a single machine on different
    ports, and I am not using clustering weblogic. Assuming I have EJB A
    deployed on managed server 1, and EJB B deployed on managed server 2.
    I want to have EJB A to invoke EJB B. In EJB A, I guess I will
    probably create the InitialContext with the URL of managed server 2,
    then do the JNDI look up and call EJB B.
    My questions are:
    - Can weblogic handle transaction that spans two managed servers
    (non-clustering setting)?
    - Does weblogic use XA to handle transaction between managed servers?
    - Do I need to do any JTA code in order to achieve that (instead of
    just letting the EJB container to handle the transaction for me)?
    Thanks in advance!
    B.L.

    Hi,
    "benson" <[email protected]> wrote in message
    news:[email protected]..
    (Weblogic 6.1 - non-clustering version)
    I have two managed servers configured on a single machine on different
    ports, and I am not using clustering weblogic. Assuming I have EJB A
    deployed on managed server 1, and EJB B deployed on managed server 2.
    I want to have EJB A to invoke EJB B. In EJB A, I guess I will
    probably create the InitialContext with the URL of managed server 2,
    then do the JNDI look up and call EJB B.
    My questions are:
    - Can weblogic handle transaction that spans two managed servers
    (non-clustering setting)?Yes, it can.
    - Does weblogic use XA to handle transaction between managed servers?Yes, it does. Make sure you use TX DataSources. If the servers are connected
    to different databases, TX DataSources should be based on connection pools
    used XA drivers.
    - Do I need to do any JTA code in order to achieve that (instead of
    just letting the EJB container to handle the transaction for me)?No, you don't. WebLogic will take care about handling
    distributed TXs.
    Regards,
    Slava Imeshev

  • Howto Programatically discover clusters, admin server, managed servers

    Hi,
              I'm want to programatically discover what clusters, admins servers, and managed servers exist in my domain.
              I've tried using get MBeanHome, but in order to get that I must know the location of the admin server.
              I'd like to be able to discover from within my program the host name and port of the admin server,and then from there to be able to discover the managed servers and clusters that are running and on what machines/ ports they are running on.
              Is this possible.
              Cheers,
              Keith

    Keith Hyland <> writes:
              > I'm want to programatically discover what clusters, admins servers, and managed servers exist in my domain.
              >
              > I've tried using get MBeanHome, but in order to get that I must know the location of the admin server.
              >
              > I'd like to be able to discover from within my program the host name and port of the admin server,and then from there to be able to discover the managed servers and clusters that are running and on what machines/ ports they are running on.
              >
              > Is this possible.
              Its possible, but there are no public APIs for directly doing this. I
              guess you could try looking up the local server's MBean and navigating
              to the admin server's MBean from there.
              You could probably also deduce it from the system property
              weblogic.management.host that you would have set to boot the managed
              server.
              andy
              --

  • Clustered enviroments -Starting/Stopping Managed servers, NodeManager & Admin server from single node

    Hi,
    We have 5 node Weblogic cluster. At the time of some mass activities like patching we have to stop all our managed, Admin Servers and node manager and it takes a lot of time to do this by logging in to all 5 managed servers and stopping all components. Though we have node manager configured so we can stop managed servers from console itself, but still have to login to each server to stop NodeManager and Oracle Http Servers.
    Things become worse when wee have to shutdown almost 15-20 instances.
    Is there a way (using wlst or shell script) that we can do all these tasks from by logging in to just one server?
    Thanks,
    Suraj

    The simple code below would shutdown all the running Servers. Some of the APIs for ex MBeanHome are deprecated..still work though..if your concerned about that..I am working on the same using JMX API..if you want I could share that too. And also as described in the below link..you can Cluster stop and start using WLST commands.  Managing the Server Life Cycle - 12c Release 1 (12.1.1)
    <This code is from Oracle Weblogic documentation>
    import java.util.Set;
    import java.util.Iterator;
    import java.rmi.RemoteException;
    import javax.naming.Context;
    import javax.management.ObjectName;
    import java.io.*;
    import java.net.*;
    import weblogic.jndi.Environment;
    import weblogic.management.MBeanHome;
    import weblogic.management.WebLogicMBean;
    import weblogic.management.configuration.ServerMBean;
    import weblogic.management.runtime.ServerRuntimeMBean;
    import weblogic.management.runtime.ServerStates;
    import weblogic.management.WebLogicObjectName;
    public class ServerStopper {
      public static void stop() throws Exception {
        MBeanHome home = null;
        //url of the Admin server
        String url = "t3://localhost:7001";
        String username = "weblogic";
        String password = "welcome1";
        ServerRuntimeMBean serverRuntime = null;
        Set mbeanSet = null;
        Iterator mbeanIterator = null;
        try {
          // Set ContextClassloader to prevent assertions
          URL[] urls = { new File("/").toURL() };
          Thread.currentThread().setContextClassLoader(new
             URLClassLoader(urls));
          Environment env = new Environment();
          env.setProviderUrl(url);
          env.setSecurityPrincipal(username);
          env.setSecurityCredentials(password);
          Context ctx = env.getInitialContext();
          home = (MBeanHome)
               ctx.lookup("weblogic.management.adminhome");
          mbeanSet = home.getMBeansByType("ServerRuntime");
          mbeanIterator = mbeanSet.iterator();
          while(mbeanIterator.hasNext()) {
            serverRuntime = (ServerRuntimeMBean)mbeanIterator.next();
            if(serverRuntime.getState().equals(ServerStates.RUNNING)){
                serverRuntime.shutdown();
        } catch (Exception e) {
          e.printStackTrace();

  • Two Managed Servers non-clustered problem

    I have two managed servers each exactly the same just listening on different
              port in the same domain. When I start the second server I get a "Child
              exited " message reported in the first server's logs. This does not seem to
              affect the running of the server. Does anyone know why this happens. Note
              that both servers define the same jndi name for a stateless session bean.
              Could this be a problem>
              

    Solaris 8. There is no information in any log relating to the message "Child
              Exited". Note that sometimes the message appears twice.
              "Kumar Allamraju" <[email protected]> wrote in message
              news:[email protected]..
              > what OS is this on?
              > This message may not be coming from WLS?
              > Can you post the error message or stacktrace?
              >
              > --
              > Kumar
              >
              > Paul
              >
              > > I have two managed servers each exactly the same just listening on
              different
              > > port in the same domain. When I start the second server I get a "Child
              > > exited " message reported in the first server's logs. This does not seem
              to
              > > affect the running of the server. Does anyone know why this happens.
              Note
              > > that both servers define the same jndi name for a stateless session
              bean.
              > > Could this be a problem>
              > >
              > >
              >
              

  • WebLogic 10.3.4 Adminserver Performance Issues - Clustered Environment.

    I am running WL 10.3.4 64 bit version on 64 bit hardware (blade serves) running Solaris 10 and I have the servers setup in a cluster. There are 3 machines are part of the cluster.
    Managed Server1 is on Blade Server 1
    Managed Server2 is on Blade Server 2
    Managed Server3 is on Blade Server 3
    We have about 20 different clusters setup and 60 servers across those 20 clusters.
    The adminserver runs on a separate box (not a blade server). Tasks that do NOT involve the servers or deployments tabs are relatively fast however, when I go to the servers or deployments tab, and to some extent the clusters tab, the response is very slow.
    For example, I am running WL 10.3.0 on less robust hardware. I installed the patch that Oracle published to resolve this particular issue. We have a similar number of clusters and servers on in that environment (3 severs to a cluster). The servers tab responds in 10 seconds.
    In WL 10.3.4, the servers and deployments tabs take about 45 - 60 seconds to respond. This adds a LOT of overhead to doing any server or deployment related tasks.
    I realize that the admin server has to query all the managed servers but performance in 10.3.4 is much worse than WL 10.3.0 with the performance patch.
    Has anyone else run into similar performance issues with the WL 10.3.4 ADMIN Console running in a clustered environment? If so, does anyone have any suggestions to help improve performance.
    I also run a WL 10.3.4 instance in a single server environment and the servers and deployments tabs respond withing 10 seconds (very fast). It is just in a clustered environment.
    Regards,
    Mark

    hi Mark, it's a very interesting story, any follow-up? I would be curious to know the reason for the delay...

  • What is the Ideal Production Setup For One Admin and 4 Managed Servers

    Dear Experts
    I will be starting with production setup including one Admin server and 4 managed servers in one single domain.
    I am thinking of creating a single node environment(no clusters) as the machine has following configuration
    OS : Windows Server 2008 R2 Datacenter
    RAM : 48 GB
    System Type : 64 bit
    Processor : Intel(Xenon) 4 processors [email protected]
    Can you please let me know if this configuration would suffice for the 4 managed servers if i assign Xmx and Xms as 4096 and Heap Space as 1024 to all the Managed Servers.
    It is very urgent and i need to convey to the Infrastructure team if harware procurement is required.
    We are looking at somewhere around 300 concurrent users(maximum load) and 100(minimum load) at a given point of time.
    Please reply ASAP.
    Thanks in advance
    Edited by: Abhinav Mittal on Apr 23, 2013 7:58 PM
    Edited by: Abhinav Mittal on Apr 23, 2013 8:03 PM

    Heap size must be calculated according to the applications that are been deployed on each JVM.
    With no deployments, you dont need more than 256k for managed servers heap size and 512k for adminserver. As biggest its your heap size, longer will take your garbage collection. And if you can prevent it, better do it.
    Kinds,
    Gabriel Abelha

  • How many managed Servers do you need?

    Hi,
    One thing that has come up time and time again, is when trying to think about a new project and a new WL domain, is trying to know how many WL Managed Servers do you need?
    Does BEA have some sort of metrics which says ok for this amount of input we recommend you having 4 managed servers on 4 single boxes?
    And is there think on have 2 managed servers on 1 box or 2 managed servers on two boxes?
    Any Help Appreciated.
    Kind Regards,
    Alistair.

    Hi :-D
    I don't believe BEA have guide lines on how many Managed servers are required based on Metrics. You could look at some performance evaluations but it's really only relevant if your application acts in the same way.
    Cheers,
    Doug.

  • Shutdown(block='false') only works via Admin Server, not managed servers

    Hello,
    We want to shut down all 10 managed servers in our domain gracefully, and in parallel. 5 of those managed servers are on the same OS/host as the AdminServer, the other 5 are on a separate OS/host (i.e. we have 5 x 2-node clusters).
    I'm writing scripts to gracefully shut down all of our servers upon an OS reboot command. On the OS with the AdminServer (and 5 managed servers), I can simply do this:
    connect(user, pwd, http://localhost:7001)
    shutdown('myserver1', 'Server', block='false')
    shutdown('myserver3', 'Server', block='false')
    shutdown('myserver5', 'Server', block='false')
    shutdown('myserver7', 'Server', block='false')
    shutdown('myserver9', 'Server', block='false')
    All 5 servers will be shutting down at the same time. If they each took 2 minutes to shut down, the whole shutdown process would take 2 minutes.
    However, on the OS running the other managed servers, I cannot assume that the AdminServer [on the other OS] will be online. Hence, I would like to gracefully shut down each server locally like this:
    connect(user, pwd, url='t3://localhost:8001')
    shutdown(block='false')
    disconnect()
    connect(user, pwd, url='t3://localhost:9001')
    shutdown(block='false')
    disconnect()
    connect(user, pwd, url='t3://localhost:10001')
    shutdown(block='false')
    disconnect()
    connect(user, pwd, url='t3://localhost:11001')
    shutdown(block='false')
    disconnect()
    connect(user, pwd, url='t3://localhost:12001')
    shutdown(block='false')
    disconnect()
    NOTE that the ports are different (i.e. I connect via WLST to the managed servers themselves rather than to the AdminServer).
    Unfortunately, in this scenario, the block='false' does not work. WLST waits until the managed server is shut down before proceeding to the next connect() command. So if each server took 2 minutes to shut down, the whole shutdown process now takes 10 minutes.
    We don't want to use the nmKill() command on the local Node Manager because we want a graceful shutdown.
    What options do we have to issue a graceful shutdown command either to a local Node Manager or to the managed servers themselves (since we can't expect the AdminServer to be online when our script runs)?
    Best regards,
    Michael

    Hi Mike,
    WLST will create it temp directory with /var/tmp/wlstTemp which is shared by all users. Since we cannot differentiated by different users which cause block on particular user at that time.
    One solution is to grant write access to all users to the directory where the WLST temporary directory is created (e.g. on Solaris /var/tmp; you can verify the default on your system by executing java utils.getProperty and search for java.io.tmpdir).
    If for whatever reason you cannot grant these access modes on the temporary directory, then you can create a directory somewhere on the file system where every user has the correct permissions.
    Start the scripting tool with one the following options to redirect the cache files to the specified directory. Dependent on your environment one or the other will apply.
    java -Djava.io.tmpdir=<path-to-tmpDir> weblogic.WLST
    or
    java -Dpython.cachedir=<path-to-tmpDir> weblogic.WLST
    This will be useful to make shutdown of servers without any problem.
    Note: block=false will always works it should not be the problem.
    Regards,
    Kal

  • Re: Precompiling JSP with admin/managed servers

    Thanks, but I'm not doing any copying.
              The admin/managed-server communication copies things to the managed server,
              which then always recompiles the pages when hit.
              -Greg
              Check out my WebLogic 6.1 Workbook for O'Reilly EJB Third Edition
              www.oreilly.com/catalog/entjbeans3 or www.titan-books.com
              "Robert Coonrad" <[email protected]> wrote in message
              news:[email protected]...
              >
              > check out post 8366...i found that i was not preserving
              > the lastmodified date on my jsps and this was causing
              > unnecessary re-compilation.
              >
              > hope it helps...
              > bobc
              >
              > "Greg Nyberg" <greg.nyberg.at.objectpartners.com> wrote:
              > >I believe I have exhausted all permutations of EARing/notEARing,
              > >WARing/notWARing, placing precompiled jsp class files in WEB-INF/classes,
              > >placing them in a static location and setting workingDir to that
              location,
              > >combinations of the above.
              > >
              > >No matter what, the managed server re-compiles pages the first time they
              > >are
              > >hit. Non admin/managed-server I have no problems.
              > >
              > >Can anyone from BEA comment on this problem? Or give me a workaround
              > >for
              > >getting a cluster working with precompiled jsps?
              > >
              > >-Greg
              > >
              > >"Greg Nyberg" <greg.nyberg.at.objectpartners.com> wrote in message
              > >news:[email protected]...
              > >> Grrr... The JSP engine is extremely frustrating! I've spent many hours
              > >> fighting the "staleness" checker in WL. I've been through all of the
              > >> newsgroup messages pertaining to pre-compiling, etc., and I've gotten
              > >> pre-compilation working on single-server deployments, but admin/managed
              > >> server deployments have me beat.
              > >>
              > >> WL6.1, SP1, Solaris
              > >>
              > >> I've done the pageCheckSeconds=-1 and the workingDir is set to a fixed
              > >> place. The fixed place contains pre-compiled versions of all jsps
              > >made on
              > >> that machine using jspc not 20 minutes earlier using the JSP files
              > >in the
              > >> exploded EAR file used by the admin server as the model for managed
              > >> servers.. The managed servers are on the same machine.
              > >>
              > >> When the admin server gives an application to a managed server, the
              > >managed
              > >> server creates a temporary directory containing all of the webapp
              > >> components, etc. The file timestamps on these files is the set by
              > >the
              > >> copying process to the time of the managed server boot (why?!?!????!?),
              > >so
              > >> the staleness check always thinks they are new and could care less
              > >what
              > >> precompiled jsps I have in my workingDir, the WEB-INF/classes
              directory,
              > >or
              > >> anywhere else. The pageCheckSeconds=-1 seems to be completely ignored
              > >in
              > >> this scenario.
              > >>
              > >> If I tell the managed server to precompile everything on boot (about
              > >45
              > >> minutes for this app) it will create versions of the classes that match
              > >th
              > >e
              > >> new JSP file timestamps, but this does not even survive a reboot of
              > >the
              > >> managed server because it AGAIN creates a new temp version of
              everything
              > >on
              > >> the next reboot with new timestamps.
              > >>
              > >> If I wait for the managed server to boot and find the directory like
              > >> .../applications/.wlnotdelete_man1/wlap7336/webapp/... and physically
              > >copy
              > >> (via cp -pr to retain timestamps) all of the original webapp components
              > >on
              > >> top of the temp versions, the staleness checker is happy and the
              > >> pre-compiled versions work fine.
              > >>
              > >> There HAS to be a way to package pre-compiled versions of the JSPs
              > >in the
              > >> "model" application in the admin server and keep from having to
              precompile
              > >> the JSPs on every managed server every time managed server is booted..
              > >>
              > >> It would help if we had a way to bypass the staleness checking
              > >completely..
              > >> Or you guys should make the timestamps on the files copied by the
              > >> admin/managed deployment process match properly so the staleness
              checker
              > >> doesn't think the JSP is different.
              > >>
              > >> It would also help if the engineer who wrote this could explain the
              > >rules
              > >> being implemented by the staleness checker. So far all the messages
              > >in
              > >the
              > >> newsgroup have amounted to point solutions for problems without a good
              > >> understanding of what the engine is checking for and/or doing under
              > >the
              > >> covers. Looking at the generated .java files for the JSP pages helps,
              > >but
              > >> it is not good enough...
              > >>
              > >> Anyone out there have a working admin/managed server JSP application?
              > >> -Greg
              > >>
              > >> -----------------------------------------------------------
              > >> Check out my WebLogic 6.1 Workbook for O'Reilly EJB Third Edition
              > >> www.oreilly.com/catalog/entjbeans3 or www.titan-books.com
              > >>
              > >>
              > >>
              > >
              > >
              >
              

    The admin/managed-server communication copies things to the managed server, which then always recompiles the pages when hit.
              This is a known issue and is fixed. The timestamps of the compiled classes was not being preserved when extracted from the war file used to distribute to the managed servers. This will be available in WLS6.1 Service Pack 3 - and there is a temporary patch available for SP2. Please ask your friendly BEA support person for it (you can refer to CR058946)
              I'd give you the patch myself, but they like to keep track of these things...
              Regards,
              Alex
              "Girish" <[email protected]> wrote in message news:[email protected]...
              >
              > "Aditya Kiran Gavvala" <[email protected]> wrote:
              > >Greg,
              > >
              > >I have been following your posts, because our application deployment
              > >ran
              > >into exact same problem you ran into. I had spent a full two days
              > >researching into the problem. And I figured the solution. Hope this
              > >helps.
              > >
              > >Here are my discoveries:
              > >
              > >The following applies only to the following environment:
              > >OS: Linux (perhaps for Win/Unix/Solaris etc)
              > >WLS 6.0 SP2 ( no rolling patches): I found Rolling Patch2 (RP2) not useful
              > >for this problem.
              > >Clustered environment with Admin/Managed servers
              > >
              > >- When you compile JSP using weblogic.jspc compiler it puts the JSP file
              > >timestamp into the compiled class. You can see it in the generated java
              > >file
              > >(you need to supply -keepgenerated switch to jspc)
              > >
              > >- When a request is made to a JSP page after the application is deployed,
              > >it
              > >seems to be retrieving this timestamp from the compiled class file and
              > >comparing it with the JSP file timestamp. If they dont match a compile
              > >command gets run by the server. Thereby you see a compile happening at
              > >run
              > >time.
              > >
              > >- If you have exploded directory deployment, when you start the managed
              > >servers they create a ".war" file (under some temp dir) with all the
              > >JSP
              > >source files going into the file. You can notice this by looking into
              > >the
              > >server log file. Therefore all JSP source files get a brand new timestamp
              > >in
              > >the archive (a timestamp later than what was put class files by
              > >weblogic.jspc). So, the server at run time sees that the timestamp in
              > >the
              > >class file is older than the JSP source file and runs a recompile. So
              > >DONT
              > >DO EXPLODED directory deployment if your environment is as described
              > >in this
              > >post.
              > >
              > >- If you have ".war" file deployment, you will not have a problem. At
              > >the
              > >start up time managed server still creates "".war" file under a temp
              > >directory however it seems to be copying the content of the your ".war"
              > >file. So, the timestamps of JSP remain the same as they were before.
              > >SO NO
              > >RE-COMPILATION.
              > >
              > >- Another important thing to remember is to make sure you specify the
              > >workingDir in the weblogic.xml file. That is where the precompiled class
              > >files should reside. This should be any directory the server uses as
              > >scratch
              > >pad to compile classes or find (pre)compiled classes. This is not a
              > >directory inside your .war file is what I am trying to get at.
              > >
              > >Hope this helps,
              > >Aditya
              > >
              > >"Greg Nyberg" <greg.nyberg.at.objectpartners.com> wrote in message
              > >news:[email protected]...
              > >> Thanks, but I'm not doing any copying.
              > >>
              > >> The admin/managed-server communication copies things to the managed
              > >server,
              > >> which then always recompiles the pages when hit.
              > >>
              > >> -Greg
              > >>
              > >> -----------------------------------------------------------
              > >> Check out my WebLogic 6.1 Workbook for O'Reilly EJB Third Edition
              > >> www.oreilly.com/catalog/entjbeans3 or www.titan-books.com
              > >>
              > >> "Robert Coonrad" <[email protected]> wrote in message
              > >> news:[email protected]...
              > >> >
              > >> > check out post 8366...i found that i was not preserving
              > >> > the lastmodified date on my jsps and this was causing
              > >> > unnecessary re-compilation.
              > >> >
              > >> > hope it helps...
              > >> > bobc
              > >> >
              > >> > "Greg Nyberg" <greg.nyberg.at.objectpartners.com> wrote:
              > >> > >I believe I have exhausted all permutations of EARing/notEARing,
              > >> > >WARing/notWARing, placing precompiled jsp class files in
              > >WEB-INF/classes,
              > >> > >placing them in a static location and setting workingDir to that
              > >> location,
              > >> > >combinations of the above.
              > >> > >
              > >> > >No matter what, the managed server re-compiles pages the first time
              > >they
              > >> > >are
              > >> > >hit. Non admin/managed-server I have no problems.
              > >> > >
              > >> > >Can anyone from BEA comment on this problem? Or give me a workaround
              > >> > >for
              > >> > >getting a cluster working with precompiled jsps?
              > >> > >
              > >> > >-Greg
              > >> > >
              > >> > >"Greg Nyberg" <greg.nyberg.at.objectpartners.com> wrote in message
              > >> > >news:[email protected]...
              > >> > >> Grrr... The JSP engine is extremely frustrating! I've spent many
              > >hours
              > >> > >> fighting the "staleness" checker in WL. I've been through all
              > >of the
              > >> > >> newsgroup messages pertaining to pre-compiling, etc., and I've
              > >gotten
              > >> > >> pre-compilation working on single-server deployments, but
              > >admin/managed
              > >> > >> server deployments have me beat.
              > >> > >>
              > >> > >> WL6.1, SP1, Solaris
              > >> > >>
              > >> > >> I've done the pageCheckSeconds=-1 and the workingDir is set to
              > >a
              > >fixed
              > >> > >> place. The fixed place contains pre-compiled versions of all
              > >jsps
              > >> > >made on
              > >> > >> that machine using jspc not 20 minutes earlier using the JSP files
              > >> > >in the
              > >> > >> exploded EAR file used by the admin server as the model for managed
              > >> > >> servers.. The managed servers are on the same machine.
              > >> > >>
              > >> > >> When the admin server gives an application to a managed server,
              > >the
              > >> > >managed
              > >> > >> server creates a temporary directory containing all of the webapp
              > >> > >> components, etc. The file timestamps on these files is the set
              > >by
              > >> > >the
              > >> > >> copying process to the time of the managed server boot
              > >(why?!?!????!?),
              > >> > >so
              > >> > >> the staleness check always thinks they are new and could care
              > >less
              > >> > >what
              > >> > >> precompiled jsps I have in my workingDir, the WEB-INF/classes
              > >> directory,
              > >> > >or
              > >> > >> anywhere else. The pageCheckSeconds=-1 seems to be completely
              > >ignored
              > >> > >in
              > >> > >> this scenario.
              > >> > >>
              > >> > >> If I tell the managed server to precompile everything on boot
              > >(about
              > >> > >45
              > >> > >> minutes for this app) it will create versions of the classes that
              > >match
              > >> > >th
              > >> > >e
              > >> > >> new JSP file timestamps, but this does not even survive a reboot
              > >of
              > >> > >the
              > >> > >> managed server because it AGAIN creates a new temp version of
              > >> everything
              > >> > >on
              > >> > >> the next reboot with new timestamps.
              > >> > >>
              > >> > >> If I wait for the managed server to boot and find the directory
              > >like
              > >> > >> .../applications/.wlnotdelete_man1/wlap7336/webapp/... and physically
              > >> > >copy
              > >> > >> (via cp -pr to retain timestamps) all of the original webapp
              > >components
              > >> > >on
              > >> > >> top of the temp versions, the staleness checker is happy and the
              > >> > >> pre-compiled versions work fine.
              > >> > >>
              > >> > >> There HAS to be a way to package pre-compiled versions of the
              > >JSPs
              > >> > >in the
              > >> > >> "model" application in the admin server and keep from having to
              > >> precompile
              > >> > >> the JSPs on every managed server every time managed server is
              > >booted..
              > >> > >>
              > >> > >> It would help if we had a way to bypass the staleness checking
              > >> > >completely..
              > >> > >> Or you guys should make the timestamps on the files copied by
              > >the
              > >> > >> admin/managed deployment process match properly so the staleness
              > >> checker
              > >> > >> doesn't think the JSP is different.
              > >> > >>
              > >> > >> It would also help if the engineer who wrote this could explain
              > >the
              > >> > >rules
              > >> > >> being implemented by the staleness checker. So far all the messages
              > >> > >in
              > >> > >the
              > >> > >> newsgroup have amounted to point solutions for problems without
              > >a
              > >good
              > >> > >> understanding of what the engine is checking for and/or doing
              > >under
              > >> > >the
              > >> > >> covers. Looking at the generated .java files for the JSP pages
              > >helps,
              > >> > >but
              > >> > >> it is not good enough...
              > >> > >>
              > >> > >> Anyone out there have a working admin/managed server JSP application?
              > >> > >> -Greg
              > >> > >>
              > >> > >> -----------------------------------------------------------
              > >> > >> Check out my WebLogic 6.1 Workbook for O'Reilly EJB Third Edition
              > >> > >> www.oreilly.com/catalog/entjbeans3 or www.titan-books.com
              > >> > >>
              > >> > >>
              > >> > >>
              > >> > >
              > >> > >
              > >> >
              > >>
              > >>
              > >
              > >
              >
              [att1.html]
              

  • Multicast Error When Starting Managed Servers

    Environment: HP-UX 11i, Weblogic 8.1 SP1.
              We're getting the following error when we start managed servers in our clustered environment:
              <Feb 24, 2004 6:33:12 PM CST> <Error> <Cluster> <BEA-000116> <Failed to join cluster PDICluster at address 2
              37.0.0.1 due to: java.net.SocketException: Can't assign requested address.
              java.net.SocketException: Can't assign requested address
              at java.net.PlainDatagramSocketImpl.join(Native Method)
              at java.net.PlainDatagramSocketImpl.join(PlainDatagramSocketImpl.java:134)
              at java.net.MulticastSocket.joinGroup(MulticastSocket.java:274)
              at weblogic.cluster.FragmentSocket.initializeMulticastSocket(FragmentSocket.java:92)
              at weblogic.cluster.FragmentSocket.start(FragmentSocket.java:109)
              at weblogic.cluster.MulticastManager.startListening(MulticastManager.java:164)
              at weblogic.cluster.ClusterCommunicationService.initialize(ClusterCommunicationService.java:52)
              at weblogic.t3.srvr.T3Srvr.initializeHere(T3Srvr.java:922)
              at weblogic.t3.srvr.T3Srvr.initialize(T3Srvr.java:664)
              at weblogic.t3.srvr.T3Srvr.run(T3Srvr.java:342)
              at weblogic.Server.main(Server.java:32)
              >
              We also get the same error when we try the util.multicastTest application. So far, we've done the following things to troubleshoot this:
              1) Different multicast addresses/ports;
              2) The util.multicastTest app.
              3) We're researching whether or machine or network may be causing this, but since the admin. server and managed servers are running on the same server, I don't see how the network could be causing this.
              Can anyone give me any ideas as to what might be going wrong? We've set this up on other machines before with the same OS and same version of Weblogic. So I'm at a loss at the moment as to what might be going wrong.
              Thanks in advance for any help...
              Brian
              

    Environment: HP-UX 11i, Weblogic 8.1 SP1.
              We're getting the following error when we start managed servers in our clustered environment:
              <Feb 24, 2004 6:33:12 PM CST> <Error> <Cluster> <BEA-000116> <Failed to join cluster PDICluster at address 2
              37.0.0.1 due to: java.net.SocketException: Can't assign requested address.
              java.net.SocketException: Can't assign requested address
              at java.net.PlainDatagramSocketImpl.join(Native Method)
              at java.net.PlainDatagramSocketImpl.join(PlainDatagramSocketImpl.java:134)
              at java.net.MulticastSocket.joinGroup(MulticastSocket.java:274)
              at weblogic.cluster.FragmentSocket.initializeMulticastSocket(FragmentSocket.java:92)
              at weblogic.cluster.FragmentSocket.start(FragmentSocket.java:109)
              at weblogic.cluster.MulticastManager.startListening(MulticastManager.java:164)
              at weblogic.cluster.ClusterCommunicationService.initialize(ClusterCommunicationService.java:52)
              at weblogic.t3.srvr.T3Srvr.initializeHere(T3Srvr.java:922)
              at weblogic.t3.srvr.T3Srvr.initialize(T3Srvr.java:664)
              at weblogic.t3.srvr.T3Srvr.run(T3Srvr.java:342)
              at weblogic.Server.main(Server.java:32)
              >
              We also get the same error when we try the util.multicastTest application. So far, we've done the following things to troubleshoot this:
              1) Different multicast addresses/ports;
              2) The util.multicastTest app.
              3) We're researching whether or machine or network may be causing this, but since the admin. server and managed servers are running on the same server, I don't see how the network could be causing this.
              Can anyone give me any ideas as to what might be going wrong? We've set this up on other machines before with the same OS and same version of Weblogic. So I'm at a loss at the moment as to what might be going wrong.
              Thanks in advance for any help...
              Brian
              

  • Problem Starting Managed Servers

    Hi,
    I am trying to start Managed Server on Sun Solaris. My admin Server and managed Server
    are on the same Machine.
    My admin Server starts up with out any problem. WhenI try to bring up my Managed
    Server its not reading the command line arguments which I supplied through the Console:
    Servers -> serverName -> Remote Start->Arguments. For some reason these command arguments
    are not read and so I am having errors like Unable to load libraries, etc.
    Also when I saw the log file for managed server the class path and the arguments
    are printed in a different way than I have set. I dont know if thats a problem or
    not. I am copying part of that for your reference.
    classpath=/opt/j2sdk1_3_1_01/jre/lib/rt.jar\:/opt/weblogic6.1\:/opt/weblogic6.1/lib/weblogic_sp.jar\:/opt/weblogic6.1/lib/weblogic.jar\
    serverStartArgs=-Dweblogic.management.server\=sunserver\:7210 -Djava.security.policy\=\=/opt/weblogic6.1/lib/weblogic.policy
    I have checked the config.xml file and what ever I have set in the console is being
    saved properly in the xml file.
    Can any one know why the back slash is printed above even though I havent typed that
    in the console. Has any one have experince with setting the command line arguments
    and class path through console for Managed servers.
    I am stuck up at this point. Any help in this regard will be greatly appreciated.
    Thanks,
    Kiran

    Seems a method signature does not match with the implementation in the
    library either due to namemangling or incompatible versions of libraries.
    You are loading your own custom native libraries for security.
    Best would be use the default startManagedServer -- once the server has
    started, copy and paste the entire value of java.library.path into the
    startup arguments for Remote start.
    For instance in my server log file:
    ####<Apr 10, 2002 6:47:10 PM PDT> <Info> <Management> <mountview> <myserver>
    <main> <system> <> <141034> <Java system property: java.library.path =
    /home/sabha/600sp2only/jdk130/jre/bin/../lib/sparc/hotspot:/home/sabha/600sp
    2only/jdk130/jre/bin/../lib/sparc::/home/sabha/600sp2only/wlserver6.0/lib/so
    laris:/home/sabha/600sp2only/wlserver6.0/lib/solaris/oci816_8:/usr/lib>
    I would set that as the arguments :
    -Djava.library.path=/home/sabha/600sp2only/jdk130/jre/lib/sparc/hotspot:/hom
    e/sabha/600sp2only/jdk130/jre/bin/../lib/sparc::/home/sabha/600sp2only/wlser
    ver6.0/lib/solaris:/home/sabha/600sp2only/wlserver6.0/lib/solaris/oci816_8:/
    usr/lib
    If this does not work, then java for some reason is not taking library paths
    set as java.library.path and I dont know how to set LD_LIBRARY_PATH as part
    of arguments to java. Another thing is before starting the NodeManager, set
    the LD_LIBRARY_PATH to point the library directories and see if this helps.
    regards,
    Sabha
    "Kiran" <[email protected]> wrote in message
    news:[email protected]...
    >
    Sabha,
    I am trying to start the managed server via remote start. After readingyour reply
    I have cleaned up the argument list for remote start. Now I am gettingaltogether
    a different error.(Surpsisingly I was able to start the managed serversusing srartManagedServer.sh.
    I dont know why I am having trouble to do the same through the remotestart).
    >
    I am getting a unsatisfied link error. In your reply you have asked me toadd -Djava.library.path
    to the argument liist which I did. Below you can find the exact error Ihave got:
    >
    >
    The WebLogic Server did not start up properly.
    Exception raised:
    java.lang.UnsatisfiedLinkError:/export/home/spbbprf1/Servers/v25_0/bb/pub/libinterface.so:
    ld.so.1: /usr/bin/../java/bin/../bin/sparc/native_threads/java: fatal:relocation
    error: file /export/home/spbbprf1/Servers/v25_0/bb/pub/libinterface.so:symbol Opsys_oneWayCryptCmpApi:
    referenced symbol not found
    at java.lang.ClassLoader$NativeLibrary.load(Native Method)
    at java.lang.ClassLoader.loadLibrary0(ClassLoader.java:1382)
    at java.lang.ClassLoader.loadLibrary(ClassLoader.java:1306)
    at java.lang.Runtime.loadLibrary0(Runtime.java:749)
    at java.lang.System.loadLibrary(System.java:820)
    atamdocs.APILink.accesscontrol.sum.EncryptApi.<clinit>(EncryptApi.java:20)
    atamdocs.APILink.accesscontrol.sum.UserJoltBackend.<clinit>(UserJoltBackend.ja
    va:52)
    >
    atamdocs.APILink.accesscontrol.sum.SumManagerImpl.<init>(SumManagerImpl.java:5
    9)
    >
    at java.lang.Class.newInstance0(Native Method)
    at java.lang.Class.newInstance(Class.java:237)
    atamdocs.uamsimpl.server.storage.multi.MultiManagerImpl.<init>(MultiManagerImp
    l.java:208)
    >
    at java.lang.Class.newInstance0(Native Method)
    at java.lang.Class.newInstance(Class.java:237)
    atamdocs.uamsimpl.server.storage.cache.CachingManagerImpl.<init>(CachingManage
    rImpl.java:81)
    >
    at java.lang.Class.newInstance0(Native Method)
    at java.lang.Class.newInstance(Class.java:237)
    at amdocs.uams.UamsSystem.getUamsManager(UamsSystem.java:156)
    atamdocs.uamsimpl.server.ejb.wl6.WLUamsDelegate.<init>(WLUamsDelegate.java:61)
    atamdocs.uamsimpl.server.ejb.wl6.WLUamsDelegate.<init>(WLUamsDelegate.java:55)
    at java.lang.Class.newInstance0(Native Method)
    at java.lang.Class.newInstance(Class.java:237)
    at amdocs.uamsimpl.server.ejb.wl6.WLUamsRealm.<init>(WLUamsRealm.java:42)
    at java.lang.Class.newInstance0(Native Method)
    at java.lang.Class.newInstance(Class.java:237)
    at weblogic.security.acl.Realm.getRealm(Realm.java:84)
    at weblogic.security.acl.Realm.getRealm(Realm.java:62)
    atweblogic.security.SecurityService.initializeRealm(SecurityService.java:258)
    at weblogic.security.SecurityService.initialize(SecurityService.java:115)
    at weblogic.t3.srvr.T3Srvr.initialize(T3Srvr.java:402)
    at weblogic.t3.srvr.T3Srvr.run(T3Srvr.java:202)
    at weblogic.Server.main(Server.java:35)
    Reason: Fatal initialization exception
    >
    >
    Please let me know if you need more description.
    Thanks,
    Kiran
    "Sabha" <[email protected]> wrote:
    Either you are attempting to start a managed server through its
    NodeManager
    for which the remote startup arguments are not correct or something else
    is
    wrong. How are you starting the server? The remote start is only when you
    want to start the server through NodeManager and not for normalinvocation
    of the startManagedServer script.
    Does the server really start? can you paste the startup arguments for the
    remote start of the managed server as saved in the config.xml.
    What is the log generated when the managed server is startedup.
    -Sabha
    "Kiran" <[email protected]> wrote in message
    news:[email protected]...
    Sabha,
    The main problem is:
    What ever I am trying to give in the command line arguments through theConsole:
    Servers -> serverName -> Remote Start->Arguments.
    is not being read during managed server start up.
    I have all the libraries defined defined in java.lib.path. I dont knowwhy
    its not
    being read by managed server during start up. Am I missing something.
    -Kiran
    "Sabha" <[email protected]> wrote:
    If its unable to load the performance packs, try setting
    java.library.path
    to the solaris lib directories. /usr/lib, /home/oracle/lib paths -- oryou
    can just copy what is in the admin weblogic.log file for the
    java.library.path (make sure the paths are correct and applicable forthe
    maanged server box also) and pass that as an argument as
    in -Djava.library.path=/usr/lib:/bea/wlserver6.1/lib/solaris...
    -Sabha
    "Kiran" <[email protected]> wrote in message
    news:[email protected]...
    Hi,
    I am trying to start Managed Server on Sun Solaris. My admin Server
    and
    managed Server
    are on the same Machine.
    My admin Server starts up with out any problem. WhenI try to bring
    up
    my
    Managed
    Server its not reading the command line arguments which I supplied
    through
    the Console:
    Servers -> serverName -> Remote Start->Arguments. For some reason
    these
    command arguments
    are not read and so I am having errors like Unable to load
    libraries,
    etc.
    Also when I saw the log file for managed server the class path and
    the
    arguments
    are printed in a different way than I have set. I dont know if thats
    a
    problem or
    not. I am copying part of that for your reference.
    classpath=/opt/j2sdk1_3_1_01/jre/lib/rt.jar\:/opt/weblogic6.1\:/opt/weblog
    i
    c
    6.1/lib/weblogic_sp.jar\:/opt/weblogic6.1/lib/weblogic.jar\
    serverStartArgs=-Dweblogic.management.server\=sunserver\:7210 -Djava.secur
    i
    t
    y.policy\=\=/opt/weblogic6.1/lib/weblogic.policy
    I have checked the config.xml file and what ever I have set in the
    console
    is being
    saved properly in the xml file.
    Can any one know why the back slash is printed above even though I
    havent
    typed that
    in the console. Has any one have experince with setting the command
    line
    arguments
    and class path through console for Managed servers.
    I am stuck up at this point. Any help in this regard will be greatlyappreciated.
    Thanks,
    Kiran

  • Administration vs Managed Servers

    We are preparing to upgrade from WLS5.1 to WLS8.1 and I am having a problem with
    the recommendation that the typical production environment have an administration
    server and one (or more) managed servers.
    In our environment, we host applications for multiple customers. Currently, each
    of these customers has a separate instance of WLS5.1 running. In many cases, the
    connection pools have the same name between instances, but with different userids/passwords.
    In the conversion to 8.1, I have been debating having a separate domain for each
    customer application versus a single domain with a separate managed server for
    each customer. The single domain has many desirable features, but it does not
    provide for the multiple userid /passwords for a connection pool. The separate
    domain per customer eliminates the connection pool issue, but essentially doubles
    the number of WLS servers (1 admin and 1 managed) for each customer.
    Even though BEA recommends a separate admin server and managed server, in reality,
    how important is this? Will I cause myself problems in the future if I choose
    to run a combined admin and managed server?
    Thanks for any thoughts on this subject.

    Thanks David. As God as my witness, the tech seriously told me that they all use the same JVM. Then I got seriously confused, like I was an idiot and didn't understand what a JVM was and I started second-guessing myself. The whole conversation was very strange and yes, there was definitely a language barrier and I'm guessing I just wasn't being clear enough...I don't know...all I know is I walked away from the phone call in worse shape then when I first opened the ticket.
    At any rate, do you have recommendations for admin server ram? I know there are performance tuning tools to help us figure out the optimum settings for our managed server, and I plan on implementing them...but going back to the ram for the Admin server, versus managed server...is there a recommended way of specifying different heap size for the managed servers? Is the recommended way to set the the -Xms and -Xmx settings via the 'Arguments' field on Environment > Servers > serverName > Configuration > Server Start tab via the admin console?
    Thanks again for your quick reply!
    Cheers,
    rlb

Maybe you are looking for