Outlook loses its connection to the exchange 2003 server 1

[drmohlen]

I was wondering if any one can help me to solve our intermittent network problem.

We have a Cisco 2970 GB switch which all servers including exchange server 2003 are plugged into and also have 3 Cisco 2950 48 port which they all linked up using cross over cable and clients machines are plugged into.

The problem is every so often the outlook on the client loses its connection to the exchange server intermittently for one or two minutes but it manages to restore it immidiatley. The problem does not happen to all users, it does happen to few users more frequent than others. We only lose connection to the exchange server and no other servers and when the connection drops I can not ping the Exchange server nor can I ping any devices from the exchange!!

It is a bizar behavior and can not put my finger on it, I have changed the NIC on the Exchange server but the problem still occurs!

Any suggestion or ideas would be greatly appreciated.

 

[markdmac]

Check the advanced settings on the NIC itself. There is an option to allow the OS to power down the NIC to conserve power. Why anyone would want to do this is beyond me and the default setting if the NIC supports this mode is for it to be enabled.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[drmohlen]

Thank you for the response, yes I was aware of that and it had been turned off, I also downloaded the latest NIC drivers from Dell website for windows 2003 server but it didn't cure the problem .any other suggestions please?

 

[pgaliardo]

Anything in the Event Viewer around the times that this happens?

 

[markdmac]

try moving the server to a different port on the switch. You may have a bad port.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[drmohlen]

No there is nothing in the event viewer arount the times it happens. I have also tried different port and also tried different switch but still no joy.

 

[markdmac]

I would check the CAT5 cables next. Make sure none are running past flouresent lights and switch out patch cables.

Also check for that power setting on the workstations too.

Last thing is make sure that if the switch has an autonegotiate that you have specified the speed you want to communicate at. If both the client and server side are set to Auto, then there can often be problems.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[drmohlen]

Okay, I am glad to hear that you also think it is a network problem rather than an exchange problem.
Can you please explain what do you mean by "switch out patch cables" in your reply?

No there isn't any foulresent lights near to the cat5 cables, previusley I also have changed the cat5 patch cables several time, sorry I should have said that in my orginal question.

The port on the cisco 2970 which the exchange uses has been set to: speed (Auto) and Duplex is set to( Auto Duplex), and the speed of the Exchange network card is also set to Auto detect, if you think this might be the problem what do I need to change them to?

 

[pgaliardo]

I would not have the switch and the NIC set to AUTO.

 

[markdmac]

The port on the cisco 2970 which the exchange uses has been set to: speed (Auto) and Duplex is set to( Auto Duplex), and the speed of the Exchange network card is also set to Auto detect, if you think this might be the problem what do I need to change them to?

The configuration you have has caused me heartache in the past. Set them both to 100 BaseT. The double Auto setting is notorious for causing negotiating problems.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[drmohlen]

Okay I will do the changes on this sunday, but before I do that I need to ask why you suggested to set them both to 100 Full duplex. The cisco switch and the NIC card on exchange are both Gigabits??

 

[markdmac]

If your cable is only CAT5 (as mentioned above) and not CAT5E you really won't be able to take advantage of the higher speed (at least not reliably). But by all means go ahead and set both to GB assuming then that you have the proper cabling and the above mention of CAT5 was an oversite.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[Echolord]

I have run into this problem, when the Exchange Server is running low on Physical Memory, therefore lots of paging to the virtual mem, which causes network traffic to back up.

With that in mind, I would still look at the cables first and work my way back to the Server.

I experienced this issue on a Dual Xeon, with 1 Gig supporting nearly 40 users. However it mainly happened to people who had 1 gig or more of email.

Hope this gives you a place to start.

 

[drmohlen]

Thank you for the advise, network wise I have tried every thing including changin CAT5 at both end client pc and smail server, I have changed the network switches on both client and server, I also changed the speed of the switch port and NIC on the mail server to Full duplex and Fixed 100MB, but unfortunatley none had fixed the problem.

In regards to your suggestion the mail server hosted on Xeon proccessor 3.2Ghz with 4GB of RAM which I have used the /3GB switch in the boot.ini to benifit from the additional memory. We have about 40 users with approximatley 40GB in total of mail storgae, one or two users have more than 1GB but the rest of the mailboxes are bellow 1GB .

The intemetent disconnection happens randomley to different users and some how it happens more frequentley at the weekend to one or two users! it is odd but true!

 

[rliebsch]

The switches cascading with crossover cable sounds like a potential botttle neck.

but aside from that...
Have you checked Disk I/O?
I recently fought for a couple months with my exchange server. I had RAID5 running across 4 disks. 80GB of DB. 50 users...

Disk I/O ended up the culprit. I can elaborate, but its around if you search on it. If you want more detail, lemmeno

Robert Liebsch
Stone Yamashita Partners

http://www.stoneyamashita.com/

 

[drmohlen]

You wrote "The switches cascading with crossover cable sounds like a potential botttle neck"

I am sorry but can you please explain what do you mean by that, and how the switches should be connected?

 

[markdmac]

Ideally you would want higher end switches that allow their backplanes to be connected so they become one virtual switch rather than cascading. Or have a switch large enough to accommodate all the needed ports.

As you have it, each switch needs to pass traffic on to another and then another etc. This causes latency.

I hope you find this post helpful.

Regards,

Mark

Check out my scripting solutions at

http://www.thespidersparlor.com/vbscript

 

[drmohlen]

Problem solved!!!!
Just to let every one knows that the outlook connectivity problem was caused by the enabled PROXY ARP in the pix for Inside interface. As soon as I disabled the Proxy ARP for the inside interface the problem had disappeard for good!! so Thank you to me!!