016100 - VERY slow performance with 2000 server

  • Thread starter Thread starter jim mcgrath
  • Start date Start date
J

jim mcgrath

I recently had 2 new switches installed at my site and
started having strange problems occure. Users (who use
roaming profiles) went from 5-10 seconds to login, to
about 5 MINUTES. When I try to access User Manager (from
NT4 server tools), it takes upwards of 5 minutes to load
the user list, where it used to take 2-3 seconds. I am
getting random "windows delayed write" errors on a 2nd
server (NTS2), and network drives disappear momentarily
and then come back. It takes forever to copy a 2mb file
to the network drive on the primary server (NTS1), it
will eventually stop and give a "path is too deep"
error. Yesterday I sat with our communications person
and tested the switch for 2 hours, which he indicated has
NO problems. I have been checking through technet, but
since I am not exactly what is happening, I am finding it
hard to find anything really relevant. Except for
KB135464 (browsing errors with UDP forwarding), these
have increased GREATLY since the switch was installed.
The switch installer (a different guy) said that the new
ones were on VLAN2 instead of the current VLAN1 the rest
were on. I have nothing to do with switch management,
but am wondering if this may have something to do with
it? He indicated that the others (on VLAN1) weren't
being changed. Plus the problem seems to be
intermittent. This morning it was fine, then stopped
working, then was fine again, then stopped working. I
have gone through the events logs on the servers and
can't see anything that's indicating a problem, other
than that above, where I am getting an event ID of 8003
repeatedly.

HELP!
 
have someone make sure that the switch ports are not set to auto. I have
seen lots and lots of problems with managed switches auto negotiating with
clients.

hardcode the ports on the swtiches to 100 Full, and make sure that the NIC
on each client (server or workstation.. doesnt matter) is also configured
for 100Full.

This should help exponentially.

NuTs
 
This was one of the first things that I had done, along
with doing 2 hours of diagnostics on the new switches to
confirm that they were configured the same as the
previous "master" switch. It has no effect at all.

j
 
what kind of swtiches are you working with ?

Jim McGrath said:
This was one of the first things that I had done, along
with doing 2 hours of diagnostics on the new switches to
confirm that they were configured the same as the
previous "master" switch. It has no effect at all.

j
Click to expand...
 
Cisco 2950 is the new master switch. The rest (3 x Cisco
2900 and 1 x Cisco 2900XL and 1 x Cisco 2950) are
uplinked to it.

j
 
Is data transfer between devices on the same switch, but transfers that need
to go from the distribution switch, up to the core switch, and down to the
other distribution switch are sucky ?

NuTs
 
All our labs are connected to their own switches (each
one has it's own). These in turn are then uplinked to
the master switch which contains the servers, printers,
my workstation (as admin) and the backbone connection
(for internet).

I reinstalled both XPSP1 on my machine and 2KSP4 on the
server and the loading of user manager has increased to
the speed it used to be. But I am still finding that the
users that are logging in from the labs are still taking
a long time to load their roaming profiles. What used to
take 5-10 seconds now takes 4-5 minutes.

We did analysis on the new master switch, and the primary
server (NTS1) seems to be getting packets from my
machine, but it takes a VERY long time to respond, like
20+ seconds in some cases.

I am still in the process of doing some testing (most
users profiles are different sizes, and this mornings
user may just have a large profile (20mb) though, it
should not take 4-5 minutes to load it. I am finding a
performance increase from my machine, but that's not to
say it's fixed everywhere. What I don't understand is
that I neither installed any new drivers/software onto
either my machine or the server to cause something like
this to happen.

j
 
Ok..

Is inter-switch performance acceptable ? (server to server, or workstation
to workstation)

NuTs
 
I have the same performance issues loading profiles both
from other switches as well as the master switch from a
workstation. There doesn't seem to be problems when
working server to server when on the same switch though.
I had corp IT check the switches, and they INSIST that
there is nothing wrong. But I was on VLAN1 and they said
they put the new switches on VLAN2, but I don't recall
them reprogramming the other 4 switches to the new VLAN.
I asked about it and was told they were too busy to check
it - though it's their job to manage my switches. I am
just flying blind trying to find SOME indication to where
the problem is coming from.
 
If you are using different VLANS and having these problems, its a routing
problem between the VLANS. I am not a network hardware guy, so I cant even
imagine how to instruct you to verify this. But, our environment has had
these types of problems, and our network group said it was because the
routing between the VLANS wasnt correct.

Thats about all I can add to this discussion.

NuTs
 
Back
Top