Re: Win2K domino 4.6 server slooooooowwwwww b/u

Sorry to be a network weenie, but....

What is the TCPWindowsize on the actual machine?  Is it 65536?  63K is
an odd number to see, you understand.  I'm wondering if you have an
RFC896 interaction going on, but I'm not sure.  Basically, if you have
TCP_NODELAY on (as you do in the dsm.opt file) you're agreeing to send
small packets, however they are created.  This keeps you from fully
utilizing the packet, and can cause massive network congestion.  Using
larger packets matters as a measure of your MTU size, so I'm curious
about that, too.  Can you send the values for your network connection?
I'm afraid I don't know enough W2K to tell you where to look, though.

-drs-

David Stabler, CATE/AIX, CNCE
Senior Systems Analyst

>>> CABANL AT MODOT DOT NET 08/21/02 01:04PM >>>





Zig-

this is Win2k on an ethernet network-- the switch isn't an SP switch.
But
this are the settings in the dsm.opt file right now.

I have played with resourceutilization, compression, and
largecommbuffer.
None make any difference.  The one thing that really gets me is that
even
with verbose on, I see NO files passed to the server since I kicked
off
this trace at 10:51.  This is going to a disk pool that is on a Shark,
with
nolimit on the file size.


TXNBytelimit            25600
TCPBuffsize       32
TCPWindowsize           63
TCPNodelay        YES
LARGECOMmbuffer         no
COMPression       No
*quiet
commrestartduration     30
commrestartinterval     20
RESOURceutilization     3




                      Zig Zag
                      <veritrash@YAHOO.        To:
ADSM-L AT VM.MARIST DOT EDU
                      COM>                     cc:
                      Sent by: "ADSM:          Subject:  Re: Win2K
domino 4.6 server slooooooowwwwww b/u
                      Dist Stor
                      Manager" <ADSM-
                      L AT VM.MARIST DOT EDU>


                      08/21/2002 11:14
                      AM
                      Please respond to
                      "ADSM: Dist Stor
                      Manager"






Lisa,

Check you client dsm.sys file(Notes client) see what
the
TCPWINDOWSIZE          1024
TCPBUFFSIZE            128
TXNBYTELIMIT           25600
TCPNODELAY             YES
LARGECOMMBUFFERS       YES

I get decent thruput with this, I am also on SP using
the switch..

--- Lisa Cabanas <CABANL AT MODOT DOT NET> wrote:
> This machine's weekly b/a incr is deathly slow.  The
> nic is set at 100/full
> and the Telcom guru has looked at the swtich, and
> there is nothing amiss
> there.
>
> But it takes forever... The incr kicked off Saturday
> morning, and was still
> going today, with lots of
> 08/17/2002 11:56:10AM ANS1809E Session is lost;
> initializing session reopen
> procedure.
> 08/17/2002 11:56:30AM ... successful
> 08/17/2002 03:36:37PM ANS1809E Session is lost;
> initializing session reopen
> procedure.
> 08/17/2002 03:36:57PM ... successful
> 08/17/2002 05:16:19PM ANS1809E Session is lost;
> initializing session reopen
> procedure.
> 08/17/2002 05:16:39PM ... successful
> 08/17/2002 07:10:09PM ANS1809E Session is lost;
> initializing session reopen
> procedure.
> 08/17/2002 07:10:29PM ... successful
> 08/17/2002 07:38:21PM ANS1809E Session is lost;
> initializing session reopen
> procedure.
> 08/17/2002 07:38:41PM ... successful
>
>
>
> I killed it off, and turned on some traces and see a
> bunch (1.022 x 10^26)
> of these in the trace out.  This seem like a bad
> thing to me.  Is this a
> TSM problem, or do I gather more info and tell the
> server/domino gurus to
> fix something?  (Andy, which trace flags will get me
> the most bang for my
> buck?)
>
>
> dsmem.cpp           (1900):   (Bad free list links)
>
>
> tia
>
> lisa


__________________________________________________
Do You Yahoo!?
HotJobs - Search Thousands of New Jobs
http://www.hotjobs.com