Strange Cluster Startup Timing Issues (Solaris 2.7)

Question

          When we start up our admin server for our WLS 6.0sp1 cluster (running on Solaris
          2.7), then try to start all the managed servers approximately simultaneously,
          the managed servers come up apparently normally (no errors in the logs) but no
          EJBs ever deploy.
          If we follow the same steps, but wait until each server is all the way up before
          starting the next, everything is fine.
          Has anyone else run into this?
          -daniel
          

Answer 1

Hmm, I wonder if they "cannot" or "will not" supply you the patch...
          Oh well, hopefully this won't get me in trouble! Here's the patch, stick it in your weblogic /lib
          directory or in CLASSPATH, and add the following to your startup command:
          -Dweblogic.rmi.queuesize=n
          And if you are having problems, don't tell BEA who gave this to you! :-)
          Gene
          "Aaron Oas" <[email protected]> wrote in message
          news:[email protected]...
          > Thanks for the information.
          >
          > We wrote to BEA support; here's the response (i.e. no patch is forthcoming):
          >
          > ----------------------------- begin BEA
          > response -----------------------------
          > FR: tommy
          >
          > CASE_ID_NUM: 248170
          >
          > MESSAGE:
          >
          > Daniel,
          >
          > My name is Tommy, and I'm a Developer Relations Engineer with
          > BEA Technical Support. I wanted to drop you a quick note to tell you
          > that I've picked up your support case.
          >
          > At this time engineering does not have a patch for this CR. This issue is
          > slated to be fixed in WLS 6.1 release which is scheduled for end of July.
          > At this time engineering is pushing to have 6.1 released on schedule, and
          > since
          > there is a valid work around for this issue, they will not provide a patch.
          >
          > The work around is to simply boot your servers in increments of about 30
          > seconds.
          >
          > Thanks in Advance,
          >
          > Tommy Butcher
          >
          > ----------------------------- end BEA response -----------------------------
          >
          > I guess it could be worse.
          >
          > - Aaron
          >
          >
          > "Gene Chuang" <[email protected]> wrote in message
          > news:[email protected]...
          > > Currently, weblogic_admin_rmi_queue native thread count is fixed at 3 (or
          > was it 4?). I did get a
          > > patch which allows me to set the count as a flag at startup, although it
          > would seem the longterm
          > > solution is for WL to self-manage the thread count dynamically. For
          > anyone who needs this interim
          > > solution, write to support and mention the CR.
          > >
          > > Gene
          > >
          > > "Aaron Oas" <[email protected]> wrote in message
          > > news:[email protected]...
          > > > Gene (and clusterers in general) -
          > > >
          > > > It seems odd that so few people have reported problems with concurrently
          > > > starting servers in a cluster if it is a limitation of
          > > > weblogic_admin_rmi_queue native thread count. Is it a limitation
          > imposed by
          > > > the Solaris kernel on weblogic (i.e. Solaris max threads per process or
          > per
          > > > user can't be set high enough in /etc/system)?
          > > >
          > > > I would think that for a mission-critical environment, starting managed
          > > > servers in a weblogic cluster would ideally be automated and
          > simultaneous;
          > > > rather than serial, and certainly rather than manual (log into each
          > solaris
          > > > box in turn and run the startup script by hand).
          > > >
          > > > Does anyone out there running weblogic clusters automate the startup of
          > > > their weblogic cluster as a whole (e.g. a startup script on the admin
          > server
          > > > box that cranks up all the remote managed servers as well)?
          > > >
          > > > I have ideas about how to do this without using rexec and .rhosts files,
          > but
          > > > it's not trivial (create a listener on each box wait for startup
          > command,
          > > > check for managed server process, run startup script...), and I'm
          > wondering
          > > > if:
          > > > 1) there is a well-known approach to starting up all the managed servers
          > on
          > > > remote boxes, and
          > > > 2) there is a solution to the apparent requirement that clustered
          > managed
          > > > servers can't start at the same time, even on separate boxes
          > > >
          > > > Hope there are some thoughts on this out there,
          > > >
          > > >
          > > > - Aaron
          > > > "Gene Chuang" <[email protected]> wrote in message
          > > > news:[email protected]...
          > > > > I've filed a issue, CR042545, where I cannot concurrently start more
          > than
          > > > 2 managed servers in
          > > > > Solaris 2.7 due to weblogic_admin_rmi_queue native thread count
          > > > limitations... hopefully it has
          > > > > been fixed in 6.1?
          > > > >
          > > > > Gene
          > > > >
          > > > > "Fred Bloggs" <[email protected]> wrote in message
          > > > news:[email protected]...
          > > > > >
          > > > > > When we start up our admin server for our WLS 6.0sp1 cluster
          > (running on
          > > > Solaris
          > > > > > 2.7), then try to start all the managed servers approximately
          > > > simultaneously,
          > > > > > the managed servers come up apparently normally (no errors in the
          > logs)
          > > > but no
          > > > > > EJBs ever deploy.
          > > > > >
          > > > > > If we follow the same steps, but wait until each server is all the
          > way
          > > > up before
          > > > > > starting the next, everything is fine.
          > > > > >
          > > > > > Has anyone else run into this?
          > > > > >
          > > > > > -daniel
          > > > >
          > > > >
          > > >
          > > >
          > >
          > >
          >
          >
          [cr042545.jar]