Failover cluster using BEA WL 7.0

Question

Hi there,
          since this is my first posting here, let me first describe my
          background: I have some experience in developing J2EE applications
          using various application servers, I have some knowledge in setting up a
          network and a BEA server, but mostly in Unix environments. I know a
          little bit of the concepts of clustering, but never did it myself.
          Now I have some problem understanding the posibilities with BEA WL
          clustering. What I want to do is the following:
          I have two machines with BEA WL Server, one should be the "master",
          the other a "backup". Normally, the master should recieve all
          requests, i.e. the backup is idle then. Only if the master fails, the
          backup should take over its functionality (fail-over). When the master
          is back again, the backup returns control (fail-back). The process of
          fail-over and fail-back may take some time, and I do not really need to
          transfer sessions between them (i.e., if the master fails, all sessions
          will be terminated). So this is really a very very low level of
          "high-availabilty" - I simply want to make sure that there is another
          server which can process my requests, if the other fails.
          The BEA server run on two Windows 2000 Advanced Server (yuck!), the
          whole situation looks like this:
          |
          --------+
          | Apache | (Win 2000)
          --------+
          |
          --------+
          | FW |
          --------+
          |
          ----------------+
          --+           --+
          | BEA | | BEA | (Win 2000)
          --+           --+
          ----------------+
          |
          --------+
          | FW |
          --------+
          |
          I read a couple of documents on BEA clustering, I read some postings in
          various newsgroups, but till now, I still do not know
          a) if I really need BEA clustering services here - and if yes
          b) how to use BEA clustering in this siutuation - which, when using
          the Apache plugin, does round-robin all the time, AFAIK.
          Maybe I only need to deploy my application to two servers and use some
          "dispatcher", which directs all requests to the master normally, and
          only if that is not available, to the backup. Can't this be done
          without BEA Clustering with the hardware shown above (using the MS
          Clustering Service and/or the BEA Apache plugin)?
          A second question is if there will be any problem with the FW between
          the Apache and the BEA when using the Apache plugin.
          Any help appreciated!
          Thanks in advance
          Jonas
          

Answer 1

          Hi Jonas,
          My experience with WLS clustering is just at the level of preparing a proof of
          concept and doing testing with toy applications. So, although this helped me
          understand how to use it I can not say I have any experience in a production environment.
          Having said that, here are my comments.
          As far as I have seen, you do not need to do any clustering at all. One of the
          biggies about failover is to be able to recover the session data and, for that
          matter, WLS provides several options (replicate, file, JDBC, cookies).
          However, since you do not care about preserving the session information all you
          really need to do is to create a bunch of managed servers (not clustered), the
          administrator server, and the server with the plug-in (Apache, WLS, Netscape).
          The application would be load balanced accross the managed servers by the plug-in,
          and if any of them fails the other ones will continue working OK.
          I understand you are talking about only two application servers. Please note
          that you do not need to have one of them offline for backup; you can have both
          of them serving your application, which will help your performance, and if one
          fails the whole load will go to the other one.
          You could have the plug-in in WLS in the administrator server and in this way
          you would need only three servers.
          Hope this helps.
          Ziweth
          Jonas Rathert <[email protected]> wrote:
          >Hi there,
          >
          >since this is my first posting here, let me first describe my
          >background: I have some experience in developing J2EE applications
          >using various application servers, I have some knowledge in setting up
          >a
          > network and a BEA server, but mostly in Unix environments. I know
          >a
          >little bit of the concepts of clustering, but never did it myself.
          >
          >Now I have some problem understanding the posibilities with BEA WL
          >clustering. What I want to do is the following:
          >
          >I have two machines with BEA WL Server, one should be the "master",
          >the other a "backup". Normally, the master should recieve all
          >requests, i.e. the backup is idle then. Only if the master fails, the
          >
          >backup should take over its functionality (fail-over). When the master
          >
          >is back again, the backup returns control (fail-back). The process of
          >
          >fail-over and fail-back may take some time, and I do not really need
          >to
          >transfer sessions between them (i.e., if the master fails, all sessions
          >
          >will be terminated). So this is really a very very low level of
          >"high-availabilty" - I simply want to make sure that there is another
          >
          >server which can process my requests, if the other fails.
          >
          >The BEA server run on two Windows 2000 Advanced Server (yuck!), the
          >whole situation looks like this:
          >
          >
          > |
          > --------+
          > | Apache | (Win 2000)
          > --------+
          > |
          > --------+
          > | FW |
          > --------+
          > |
          > ----------------+
          > --+           --+
          > | BEA | | BEA | (Win 2000)
          > --+           --+
          > ----------------+
          > |
          > --------+
          > | FW |
          > --------+
          > |
          >
          >I read a couple of documents on BEA clustering, I read some postings
          >in
          >various newsgroups, but till now, I still do not know
          >
          > a) if I really need BEA clustering services here - and if yes
          >
          > b) how to use BEA clustering in this siutuation - which, when using
          > the Apache plugin, does round-robin all the time, AFAIK.
          >
          >Maybe I only need to deploy my application to two servers and use some
          >
          >"dispatcher", which directs all requests to the master normally, and
          >
          >only if that is not available, to the backup. Can't this be done
          >without BEA Clustering with the hardware shown above (using the MS
          >Clustering Service and/or the BEA Apache plugin)?
          >
          >A second question is if there will be any problem with the FW between
          >
          >the Apache and the BEA when using the Apache plugin.
          >
          >Any help appreciated!
          >
          >Thanks in advance
          >
          > Jonas
          >
          >
          >
          >
          >
          >
          >
          >
          >
          >
          >
          >