Is Servlet failover supported by WebLogic5.1

Question

          Hi,
          I want to know whether servlet failover is supported
          by WebLogic5.1. To illustrate with an example:
          I have a server instance(which is part of cluster)
          which is running a servlet. If the Server instance
          fails when the servlet service method is in progress,
          will the failover happen? I mean will the secondary
          server re-invoke the service method of the servlet?
          I have tried Servlet failover with weblogic clusters
          and it simply doesnt work.
          For Object(EJB) clustering there is an option of
          declaring the Object as idempotent, but in case
          of servlets there is no such provision. Shouldn't
          Servlets be idempotent by default?
          Any ideas/comments about the same will be appreciated.
          Regards,
          Amol
          ><
          

Answer 1

          if ConnectTimeoutSecs is less than HungServerRecoverSecs
          then
          wouldn't the connection timeout occur before server fail-over
          happens?
          Viresh Garg <[email protected]> wrote:
          >
          >
          >Amol wrote:
          >
          >> Hello.
          >> Thanks Mike, Kumar and Glen for your responses.
          >> I got another query, but before that, just
          >a couple
          >> of notings from my experiences with weblogic
          >> clustering - servlet failure. (Our configuration
          >> uses iPlanet Webserver with NSAPI plugin)
          >>
          >> --- NOTINGS ---
          >> In the obj.conf file, i needed to add the
          >following
          >> parameters: HungServerRecoverSecs, ConnectRetrySecs,
          >> ConnectTimeoutSecs. Also I noted that(correct
          >me if
          >> i'm wrong) the foll relation ought to be maintained
          >> to allow Servlet failover to take place:
          >> "ConnectRetrySecs < HungServerRecoverSecs
          >< ConnectTimeoutSecs"
          >
          >There is no connection of connectretrySecs and
          >HungServerRecoverSecs.
          >
          >HungServerRecoverSecs means we connected to
          >WLS successfully and now
          >waiting for response in proxy code, if response
          >doesn't come back in
          >these many secs, we assume server is hung (
          >you can connect to it but
          >can't get a response back) and we do failover.
          >
          >ConnectRetrySecs is the time for which we sleep
          >before retrying once we
          >have tried all the servers in static and dynamic
          >list and we couldn't
          >connect to a server.
          >
          >ConnectTimeoutSecs is the time upto which we
          >will keep on sleeping and
          >trying to connect to a server in loop until
          >we can either connect, or
          >connection timeout has expired.
          >
          >Viresh Garg
          >Principal Developer Relations Engineer
          >BEA Systems
          >
          >>
          >>
          >> This is not mentioned in the docs and more
          >importantly
          >> the default values assigned to these parameters
          >do
          >> NOT satisfy the above mentioned condition
          >hence even
          >> though Servlets may be IDEMPOTENT by default
          >as Kumar
          >> points out, Servlet fail-over would not work.
          >>
          >> Also some coding change is required if one
          >is using
          >> servlet calling Beans in a transaction. We
          >need to
          >> maintain servlet-level transactions by setting
          >a flag
          >> in the session.
          >> --- END OF NOTING ---
          >>
          >> Finally the new query :)
          >> "Consider the foll scenario:
          >> Request comes to Servlet, is fwded to SessionBean
          >> which calls multiple methods on EntityBean
          >in a
          >> SINGLE transaction. Now consider the situation
          >> wherein the app server continues working fine,
          >> but network connection for the machine hosting
          >the
          >> app server is disconnected. The proxy receives
          >no
          >> response from the concerned app server and
          >fails over
          >> to the secondary server. However the first
          >app server
          >> cannot perform the update because of the lock
          >n the
          >> DB by the first app server. How does one resolve
          >such
          >> a situation?"
          >>
          >> We simulated this situation by removing the
          >network
          >> cord while the transaction was in progress...
          >and the
          >> response to the client was an ugly "Proxy
          >failed"
          >> error!
          >>
          >> Thanks and Best Regards,
          >>
          >> Amol
          >> ><
          >> Any ideas would be greatly appreciated.
          >> I think we need to
          >>
          >> glen wilcox <[email protected]> wrote:
          >> >I guess we are all assuming that one of the
          >"Bridge" products
          >> >(Netscape,Apache) is being used otherwise
          >idempotent
          >> >doesn't apply. The
          >> >failover will work but you need a proxy server
          >in front
          >> >of the cluster.
          >> >If you are using the session you will also
          >need to configure
          >> >for session
          >> >replication.
          >> >
          >> >Glen
          >> >
          >> >Kumar Allamraju wrote:
          >> >
          >> >> BTW, idempotent is "ON" by default
          >> >>
          >> >> Mike Reiche wrote:
          >> >>
          >> >> > Look for an 'idempotent' setting for
          >specific URLs.
          >> >> >
          >> >> > http://www.weblogic.com/docs51/admindocs/nsapi.html
          >> >> >
          >> >> > Mike
          >> >> >
          >> >> > "Amol" <[email protected]> wrote:
          >> >> > >
          >> >> > >Hi,
          >> >> > >I want to know whether servlet failover
          >is supported
          >> >> > >by WebLogic5.1. To illustrate with an
          >example:
          >> >> > >I have a server instance(which is part
          >of cluster)
          >> >> > >which is running a servlet. If the Server
          >instance
          >> >> > >fails *when the servlet service method
          >is in progress*,
          >> >> > >will the failover happen? I mean will
          >the secondary
          >> >> > >server re-invoke the service method
          >of the servlet?
          >> >> > >
          >> >> > >I have tried Servlet failover with weblogic
          >clusters
          >> >> > >and it simply doesnt work.
          >> >> > >
          >> >> > >For Object(EJB) clustering there is
          >an option of
          >> >> > >declaring the Object as idempotent,
          >but in case
          >> >> > >of servlets there is no such provision.
          >Shouldn't
          >> >> > >Servlets be idempotent by default?
          >> >> > >
          >> >> > >Any ideas/comments about the same will
          >be appreciated.
          >> >> > >
          >> >> > >Regards,
          >> >> > >Amol
          >> >> > >><
          >> >
          >