[SunRay-Users] Strange Failover Group behaviour

ottomeister ottomeister at gmail.com
Sat Mar 20 21:09:23 EET 2010


On Sat, Mar 20, 2010 at 4:08 AM, Craig Hummer <craig.hummer at cnt.co.uk> wrote:
> We have been having issues with one of our sites.
>
> 2 x SRSS4.2 running SRWC2.2 in a failover group.
>
> I have patched to date but am still having an issue were sessions are
> created on one server then after a time the session disconnects and a new
> one is created on the other server.  The disconnected session stays visible
> in the Login Greeter \ Idle window in the GUI.  Terminating the disconnected
> session results in (again, after a time) the connected session becoming
> disconnected and moving into the lower section and a connected session in
> the upper section.
>
> I have confirmed using utreplica that my primary and secondary are
> configured correctly and utgstatus shows that both servers are in a group
> (even though temporarily the U flag disappeared on one of them until I did a
> utrestart).

The disappearance of the 'U' flag means that the servers are failing to
communicate over that interface.  The server that stops displaying the
'U' has seen no group membership announcements from the other
server in the last 60 seconds.  If no interfaces have the 'U' flag for the
other server then this server will consider the other server to be
unavailable, so it will no longer attempt to redirect DTUs to that other
server for the purposes of session reconnection or new session
creation.

However, that shouldn't interfere with any existing DTU connections to
the second server.  If the DTU gets disconnected then that implies that
there's also a network problem between the second server and the
DTU.  It takes about a minute for a DTU to give up on a server
connection. If the DTU does give up and then makes a new connection
to the first server (the one that no longer has a 'U' path to the second
server) then that first server won't attempt to reconnect the DTU to its
existing session on the unreachable second server, it'll create a new
session for the DTU.

This all points to a network problem with the second system.  Is it
logging link errors or reporting other network issues?  Is it possible
that some other system is taking its IP address?  Is anything strange
happening to the switch it's connected to?

The disappearing 'U' can be caused by a loss of the multicast path
between the servers but that on its own wouldn't affect the DTU
connection, so I think you're looking for something more serious
than that.  (If that was the only problem you were seeing then the
workaround for intermittent multicast dropouts would be to switch
to broadcast announcements instead, or to configure unicast
announcements.)

OttoM.
__
Disclaimer: I am employed by Oracle. The statements and opinions
expressed here are my own and do not necessarily represent those
of Oracle Corporation.


More information about the SunRay-Users mailing list