Gary Mills wrote:
On Fri, Nov 02, 2007 at 02:17:46PM -0400, Bob Doolittle wrote:
What kind of hardware platform are you using?
Are you using Solaris or Linux?
Solaris 10 8/07 x86.
Hardware configuration please? How much memory?
How many sessions on your server (not just active ones)?
If you have an undersized and overloaded server,
any period of intense system activity can cause
temporary starvation of heartbeats to the Sun Ray
which can result in timeouts such as you describe
if the resource oversubscription is sufficiently
large and/or kernel scheduling is sufficiently
poor. So we need to understand your situation better.
No, it's not that. The Sun Ray server is typically idle.
I can produce the timeout reliably with those three triggers.
It only happens with my Sun Ray behind my NAT router.
I occasionally see this on small systems which are paging heavily
(e.g. a U40 with 2GB memory and 4-5 sessions). This can happen
even with only one active Sun Ray if there are a number
of inactive sessions with processes with large RSS segments which
can't be paged out.
But I don't know how NAT would come into play. There's little
in common with logging into Solaris and logging into Windows
with uttsc. Perhaps larger WAN latencies and/or a lossy WAN
link exacerbate the problem of heartbeat starvation under paging
load. What does utcapture tell you during this period? How about
vmstat, do you get a spike in paging activity during this period?
-Bob
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users
