N series, MPE 7.5 , 100BT card. Remote offices served by an MPLS connection , dedicated line. Each office has its gateway connected to a switch, then around the office to pcs running reflections, couple of printers.
Checking QOS from central shows each office gets an average of 1 event of latency>1000ms per day, sometimes in working hours, sometimes not.

Users email connects them to MS EXCh 2007 in central office, logged on to domain.

In one office only, a couple of users, each with multiple reflection sessions open, have the currently working Reflection session drop off, and Outlook comes up with a logon prompt, indicating it lost connection to its server. The other reflection sessions don't drop. At most 3 users out of 10 have reported the problem.

I am struggling to find a way to nail down the cause. checking console for msgs around the time, have seen JINETD report 'could not initialize data in path with TCP' a couple of times, but not always, so this could be a symptom of connectivity loss, and not a cause. The dropped sessions show up with netipc error 42, which makes sense - its what you get if you just kill reflections without logging off.

nettool res doesn't show any tables busted.  I doubt it's a failure of data link into the office, as this should drop all connections to everybody, so my suspicion is that's it something flaky in the internal network.

Anybody seen anything like this please, or have any ideas where to look?
thanks,
John

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *