Randomly dropping connection to LAN

With over five years installing and maintaining networks I thought I'd seen just about everything but this one has me stumped and I'm running out of places to look for solutions... Here's the short version...

Stand-alone Win2K File and Print Server, workstations running a mix of Win98SE, Win2K and XP Pro. All workstations use dynamic IP from DHCP server.

The server as well as all PCs, hubs, routers, and switches are plugged into APC UPS units. All cabling is surge protected at the hub, router or switch as applicable using APC ProtectNet devices.

The problem...

One particular workstation is randomly losing the connection to the LAN. It might happen two or three times in one day, or it could go for several weeks without a problem.

Troubleshooting steps so far include...
....New NIC
....New patch cables
....Changed ports on the switch
....Gave the workstation a static IP address
....Deleted and recreated the user account on the server
....Replaced the problem PC with a different one - the problem did not move with the PC.

Thinking it was a bad cable run, we replaced the entire CAT5e cable run from the wall jack to the switch using shielded cable - problem is still there.

Any and all help would be most appreciated.

--

Dave

I've got the same problem here. Except, all my NT4.0
client's are solid, and all my XP client's except for one
are loosing connectivity to the server (they drop off
randomly, but they seem to loose their connection after
long periods of being idle). I have tried everything,
from new hardware to tweaking to doing clean installs and
starting over with different configurations.

At this point, I think it is either DNS, or Active
Directory. The XP machine that is solid, is configured
the same as the other XP box's, I have even copied the
user account from this machine to others on the net,
nothing seems to help.

If you look around the news groups we apparently are not
alone in this, but I have yet to get any solid feed back
on potential causes.

If you do make progress, can you inform me of it! Any
help at this point would be greatly appreciated.

>-----Original Message-----
>With over five years installing and maintaining networks
I thought I'd seen just about everything but this one has
me stumped and I'm running out of places to look for
solutions... Here's the short version...
>
>Stand-alone Win2K File and Print Server, workstations
running a mix of Win98SE, Win2K and XP Pro. All
workstations use dynamic IP from DHCP server.
>
>The server as well as all PCs, hubs, routers, and
switches are plugged into APC UPS units. All cabling is
surge protected at the hub, router or switch as applicable
using APC ProtectNet devices.
>
>The problem...
>
>One particular workstation is randomly losing the
connection to the LAN. It might happen two or three times
in one day, or it could go for several weeks without a
problem.
>
>Troubleshooting steps so far include...
>....New NIC
>....New patch cables
>....Changed ports on the switch
>....Gave the workstation a static IP address
>....Deleted and recreated the user account on the server
>....Replaced the problem PC with a different one - the
problem did not move with the PC.
>
>Thinking it was a bad cable run, we replaced the entire
CAT5e cable run from the wall jack to the switch using
shielded cable - problem is still there.
>
>Any and all help would be most appreciated.
>
>--
>
>Dave
>.
>

No problem - when I finally get it figured out I'll be sure to post the solution that worked for me.

--

Dave

"Jeff" <(E-Mail Removed)> wrote in message news:0b9401c3b9eb$4f1dec00$(E-Mail Removed)...
> I've got the same problem here. Except, all my NT4.0
> client's are solid, and all my XP client's except for one
> are loosing connectivity to the server (they drop off
> randomly, but they seem to loose their connection after
> long periods of being idle). I have tried everything,
> from new hardware to tweaking to doing clean installs and
> starting over with different configurations.
>
> At this point, I think it is either DNS, or Active
> Directory. The XP machine that is solid, is configured
> the same as the other XP box's, I have even copied the
> user account from this machine to others on the net,
> nothing seems to help.
>
> If you look around the news groups we apparently are not
> alone in this, but I have yet to get any solid feed back
> on potential causes.
>
> If you do make progress, can you inform me of it! Any
> help at this point would be greatly appreciated.
>
> >-----Original Message-----
> >With over five years installing and maintaining networks
> I thought I'd seen just about everything but this one has
> me stumped and I'm running out of places to look for
> solutions... Here's the short version...
> >
> >Stand-alone Win2K File and Print Server, workstations
> running a mix of Win98SE, Win2K and XP Pro. All
> workstations use dynamic IP from DHCP server.
> >
> >The server as well as all PCs, hubs, routers, and
> switches are plugged into APC UPS units. All cabling is
> surge protected at the hub, router or switch as applicable
> using APC ProtectNet devices.
> >
> >The problem...
> >
> >One particular workstation is randomly losing the
> connection to the LAN. It might happen two or three times
> in one day, or it could go for several weeks without a
> problem.
> >
> >Troubleshooting steps so far include...
> >....New NIC
> >....New patch cables
> >....Changed ports on the switch
> >....Gave the workstation a static IP address
> >....Deleted and recreated the user account on the server
> >....Replaced the problem PC with a different one - the
> problem did not move with the PC.
> >
> >Thinking it was a bad cable run, we replaced the entire
> CAT5e cable run from the wall jack to the switch using
> shielded cable - problem is still there.
> >
> >Any and all help would be most appreciated.
> >
> >--
> >
> >Dave
> >.
> >

Assuming a hardware problems:
1) Route the NIC cable by a completely different path, such as
a temporary connection down the hall. Then see what happens.

2) Run an extension cord from same wall receptacle that the
'faulty' station used to a wall receptacle at the hub. Then
using a 3.5 digital volt meter, check voltages between that
'station' wall receptacle and voltages on the 'hub' wall
receptacle. For example, no difference greater than 2 volts
AC between each safety ground and both receptacle's neutral
wires (4 measurements). Less than 10 volts difference between
neutrals for each receptacle. IOW extension cord permits
measuring these voltages from different parts of the building
while everything else in that building is in operation.

3) Network cards have extensive diagnostics. Have seen where
every diagnostic works just fine, but then the NIC fails only
when doing long data exchanges with both NICs. Diagnostics
require that both NICs be from same manufacturer. Both NICs
can exchange test data in comprehensive testing without
interfering with the rest of the network. This test can
locate strange intermittents exactly at the time that some
external event causes the failure. To see exactly when
failure happens is also a powerful tool.

4) How is shield connected? Shield should be connected to
'chassis' ground only at one one - not at both.

5) This is also where an oscilloscope comes in handy. In one
case, saw the offending NIC cable that created rounded off
waves. Once the loose connection was solved, the NIC output
sharper (squarer) waves.

6) You are fooling yourself if you think a plug-in UPS is
doing anything more than battery backup. Plug-in UPSes
connect computers directly to AC mains when not in battery
backup mode. They are careful to word things so that some
people 'feel' they provide all kinds of additional
protection. In reality, the only thing they do reliably is
protect data from blackouts and brownouts. Many can output
quite dirty (spikey) power when in battery backup mode. No
problem since that dirty electricity is not a problem to any
properly constructed computer.

For example, this UPS in battery backup mode outputs two 200
volt square waves with a 270 volt spike between those square
waves. This is called 120 VAC. Yes. It is equivalent to 120
VAC. And since computer power supplies are so resilient, then
that very dirty backup power does not upset computers. But
when getting others to promote the myth, that same UPS
manufacturer would have you 'feel' is provides cleaner power.
It is just another reason why that oscilloscope is such a
powerful tool.

You don't need the scope. Just appreciate that the UPS is
not a solution to power problems that cause hardware
problems. Suggestion #3 is the first one I would try. It
would demonstrate whether this is a hardware or software
problem.

Dave wrote:
> No problem - when I finally get it figured out I'll be sure to
> post the solution that worked for me.
>
> Dave
>>
>>>-----Original Message-----
>>> With over five years installing and maintaining networks
>>> I thought I'd seen just about everything but this one has
>>> me stumped and I'm running out of places to look for
>>> solutions... Here's the short version...
>>>
>>> Stand-alone Win2K File and Print Server, workstations
>>> running a mix of Win98SE, Win2K and XP Pro. All
>>> workstations use dynamic IP from DHCP server.
>>>
>>> The server as well as all PCs, hubs, routers, and
>>> switches are plugged into APC UPS units. All cabling is
>>> surge protected at the hub, router or switch as applicable
>>> using APC ProtectNet devices.
>>>
>>>The problem...
>>>
>>> One particular workstation is randomly losing the
>>> connection to the LAN. It might happen two or three times
>>> in one day, or it could go for several weeks without a
>>> problem.
>>>
>>>Troubleshooting steps so far include...
>>>....New NIC
>>>....New patch cables
>>>....Changed ports on the switch
>>>....Gave the workstation a static IP address
>>>....Deleted and recreated the user account on the server
>>>....Replaced the problem PC with a different one - the
>> problem did not move with the PC.
>>>
>>> Thinking it was a bad cable run, we replaced the entire
>>> CAT5e cable run from the wall jack to the switch using
>>> shielded cable - problem is still there.
> >
>>> Any and all help would be most appreciated.

Also Just make sure that you have changed the patch cable
on both ends and changed ports on the patch panel where
the hub is. I would also make sure your not using the same
port on the hub.



>-----Original Message-----
> Assuming a hardware problems:
>1) Route the NIC cable by a completely different path,
such as
>a temporary connection down the hall. Then see what
happens.
>
>2) Run an extension cord from same wall receptacle that
the
>'faulty' station used to a wall receptacle at the hub.
Then
>using a 3.5 digital volt meter, check voltages between
that
>'station' wall receptacle and voltages on the 'hub' wall
>receptacle. For example, no difference greater than 2
volts
>AC between each safety ground and both receptacle's
neutral
>wires (4 measurements). Less than 10 volts difference
between
>neutrals for each receptacle. IOW extension cord permits
>measuring these voltages from different parts of the
building
>while everything else in that building is in operation.
>
>3) Network cards have extensive diagnostics. Have seen
where
>every diagnostic works just fine, but then the NIC fails
only
>when doing long data exchanges with both NICs.
Diagnostics
>require that both NICs be from same manufacturer. Both
NICs
>can exchange test data in comprehensive testing without
>interfering with the rest of the network. This test can
>locate strange intermittents exactly at the time that some
>external event causes the failure. To see exactly when
>failure happens is also a powerful tool.
>
>4) How is shield connected? Shield should be connected to
>'chassis' ground only at one one - not at both.
>
>5) This is also where an oscilloscope comes in handy. In
one
>case, saw the offending NIC cable that created rounded off
>waves. Once the loose connection was solved, the NIC
output
>sharper (squarer) waves.
>
>6) You are fooling yourself if you think a plug-in UPS is
>doing anything more than battery backup. Plug-in UPSes
>connect computers directly to AC mains when not in battery
>backup mode. They are careful to word things so that some
>people 'feel' they provide all kinds of additional
>protection. In reality, the only thing they do reliably
is
>protect data from blackouts and brownouts. Many can
output
>quite dirty (spikey) power when in battery backup mode.
No
>problem since that dirty electricity is not a problem to
any
>properly constructed computer.
>
> For example, this UPS in battery backup mode outputs
two 200
>volt square waves with a 270 volt spike between those
square
>waves. This is called 120 VAC. Yes. It is equivalent
to 120
>VAC. And since computer power supplies are so resilient,
then
>that very dirty backup power does not upset computers.
But
>when getting others to promote the myth, that same UPS
>manufacturer would have you 'feel' is provides cleaner
power.
>It is just another reason why that oscilloscope is such a
>powerful tool.
>
> You don't need the scope. Just appreciate that the UPS
is
>not a solution to power problems that cause hardware
>problems. Suggestion #3 is the first one I would try. It
>would demonstrate whether this is a hardware or software
>problem.
>
>Dave wrote:
>> No problem - when I finally get it figured out I'll be
sure to
>> post the solution that worked for me.
>>
>> Dave
>>>
>>>>-----Original Message-----
>>>> With over five years installing and maintaining
networks
>>>> I thought I'd seen just about everything but this one
has
>>>> me stumped and I'm running out of places to look for
>>>> solutions... Here's the short version...
>>>>
>>>> Stand-alone Win2K File and Print Server, workstations
>>>> running a mix of Win98SE, Win2K and XP Pro. All
>>>> workstations use dynamic IP from DHCP server.
>>>>
>>>> The server as well as all PCs, hubs, routers, and
>>>> switches are plugged into APC UPS units. All cabling
is
>>>> surge protected at the hub, router or switch as
applicable
>>>> using APC ProtectNet devices.
>>>>
>>>>The problem...
>>>>
>>>> One particular workstation is randomly losing the
>>>> connection to the LAN. It might happen two or three
times
>>>> in one day, or it could go for several weeks without a
>>>> problem.
>>>>
>>>>Troubleshooting steps so far include...
>>>>....New NIC
>>>>....New patch cables
>>>>....Changed ports on the switch
>>>>....Gave the workstation a static IP address
>>>>....Deleted and recreated the user account on the
server
>>>>....Replaced the problem PC with a different one - the
>>> problem did not move with the PC.
>>>>
>>>> Thinking it was a bad cable run, we replaced the
entire
>>>> CAT5e cable run from the wall jack to the switch using
>>>> shielded cable - problem is still there.
>> >
>>>> Any and all help would be most appreciated.
>.
>

Thanks for the reply and please accept my apologies for the delay in returning to this post.

With the exception of scoping the signals I think we've pretty much covered the possibilities suggested.

1) Did that - no change in symptoms
2) Outlet voltage is 121.3 VAC at the workstation; 121.2 at the hub; 121.0 at the router and 122 at the server.
3) NIC diags show no problem - all NICs in use are 3COM 3C-509C-TXM
4) Shields are grounded at one end only and all cabling meets or exceeds EIA/TIA 568B
5) We do have a quad trace scope available but after replacing the NIC in the offending workstation, and then replacing the entire workstation with a known good system, we don't think malformed signal is the cause. We also replaced the hub and router at the same time as the workstation replacement.
6) Good points all. I'm familiar with the limitations and inner workings of various UPS units having been a missile systems tech for several years in the military.

Grasping straws, I borrowed a Fluke 43B Power Quality Analyzer and left it connected to the outlet for seven days, then moved it to the output side of the UPS for another seven days. No abnormal power fluctuations were recorded on either test and the workstation still dropped off the network three times.

The only thing left that I can think of is the operating system. I have seen one instance where changing the OS solved the problem. I admit this is a long shot but that's all that's left. We'll be changing the OS on the 26th and will post the results a few days later...

--

Dave
-----------------------------------------------------------------
I wonder why evolution hasn't got around to creating an undo button ...



"w_tom" <(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
> Assuming a hardware problems:
> 1) Route the NIC cable by a completely different path, such as
> a temporary connection down the hall. Then see what happens.
>
> 2) Run an extension cord from same wall receptacle that the
> 'faulty' station used to a wall receptacle at the hub. Then
> using a 3.5 digital volt meter, check voltages between that
> 'station' wall receptacle and voltages on the 'hub' wall
> receptacle. For example, no difference greater than 2 volts
> AC between each safety ground and both receptacle's neutral
> wires (4 measurements). Less than 10 volts difference between
> neutrals for each receptacle. IOW extension cord permits
> measuring these voltages from different parts of the building
> while everything else in that building is in operation.
>
> 3) Network cards have extensive diagnostics. Have seen where
> every diagnostic works just fine, but then the NIC fails only
> when doing long data exchanges with both NICs. Diagnostics
> require that both NICs be from same manufacturer. Both NICs
> can exchange test data in comprehensive testing without
> interfering with the rest of the network. This test can
> locate strange intermittents exactly at the time that some
> external event causes the failure. To see exactly when
> failure happens is also a powerful tool.
>
> 4) How is shield connected? Shield should be connected to
> 'chassis' ground only at one one - not at both.
>
> 5) This is also where an oscilloscope comes in handy. In one
> case, saw the offending NIC cable that created rounded off
> waves. Once the loose connection was solved, the NIC output
> sharper (squarer) waves.
>
> 6) You are fooling yourself if you think a plug-in UPS is
> doing anything more than battery backup. Plug-in UPSes
> connect computers directly to AC mains when not in battery
> backup mode. They are careful to word things so that some
> people 'feel' they provide all kinds of additional
> protection. In reality, the only thing they do reliably is
> protect data from blackouts and brownouts. Many can output
> quite dirty (spikey) power when in battery backup mode. No
> problem since that dirty electricity is not a problem to any
> properly constructed computer.
>
> For example, this UPS in battery backup mode outputs two 200
> volt square waves with a 270 volt spike between those square
> waves. This is called 120 VAC. Yes. It is equivalent to 120
> VAC. And since computer power supplies are so resilient, then
> that very dirty backup power does not upset computers. But
> when getting others to promote the myth, that same UPS
> manufacturer would have you 'feel' is provides cleaner power.
> It is just another reason why that oscilloscope is such a
> powerful tool.
>
> You don't need the scope. Just appreciate that the UPS is
> not a solution to power problems that cause hardware
> problems. Suggestion #3 is the first one I would try. It
> would demonstrate whether this is a hardware or software
> problem.
>
> Dave wrote:
> > No problem - when I finally get it figured out I'll be sure to
> > post the solution that worked for me.
> >
> > Dave
> >>
> >>>-----Original Message-----
> >>> With over five years installing and maintaining networks
> >>> I thought I'd seen just about everything but this one has
> >>> me stumped and I'm running out of places to look for
> >>> solutions... Here's the short version...
> >>>
> >>> Stand-alone Win2K File and Print Server, workstations
> >>> running a mix of Win98SE, Win2K and XP Pro. All
> >>> workstations use dynamic IP from DHCP server.
> >>>
> >>> The server as well as all PCs, hubs, routers, and
> >>> switches are plugged into APC UPS units. All cabling is
> >>> surge protected at the hub, router or switch as applicable
> >>> using APC ProtectNet devices.
> >>>
> >>>The problem...
> >>>
> >>> One particular workstation is randomly losing the
> >>> connection to the LAN. It might happen two or three times
> >>> in one day, or it could go for several weeks without a
> >>> problem.
> >>>
> >>>Troubleshooting steps so far include...
> >>>....New NIC
> >>>....New patch cables
> >>>....Changed ports on the switch
> >>>....Gave the workstation a static IP address
> >>>....Deleted and recreated the user account on the server
> >>>....Replaced the problem PC with a different one - the
> >> problem did not move with the PC.
> >>>
> >>> Thinking it was a bad cable run, we replaced the entire
> >>> CAT5e cable run from the wall jack to the switch using
> >>> shielded cable - problem is still there.
> > >
> >>> Any and all help would be most appreciated.

Dave,
I have a couple questions for you.
1. What NIC's have you used that are demonstrating this problem? (other than the 3c905c-txm?)
2. Are the NIC's you have tried add-in PCI or integrated (LOM)?
3. What type of PC's are these in?
4. What OS?
5. What make / model switch/hub are you plugging into?

I am aware of certain Dell Workstation models w/ integrated 3Com NIC's that when connected to NetGear or Linksys hubs/switches can potentially demonstrate the problem you've described.

Thanks,

Doug Wyant [MSFT]
Microsoft Corporation

--
Please do not send e-mail directly to this alias. This alias is for newsgroup purposes only.

This posting is provided "AS IS" with no warranties, and confers no rights.


"Dave" <dont-(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
Thanks for the reply and please accept my apologies for the delay in returning to this post.

With the exception of scoping the signals I think we've pretty much covered the possibilities suggested.

1) Did that - no change in symptoms
2) Outlet voltage is 121.3 VAC at the workstation; 121.2 at the hub; 121.0 at the router and 122 at the server.
3) NIC diags show no problem - all NICs in use are 3COM 3C-509C-TXM
4) Shields are grounded at one end only and all cabling meets or exceeds EIA/TIA 568B
5) We do have a quad trace scope available but after replacing the NIC in the offending workstation, and then replacing the entire workstation with a known good system, we don't think malformed signal is the cause. We also replaced the hub and router at the same time as the workstation replacement.
6) Good points all. I'm familiar with the limitations and inner workings of various UPS units having been a missile systems tech for several years in the military.

Grasping straws, I borrowed a Fluke 43B Power Quality Analyzer and left it connected to the outlet for seven days, then moved it to the output side of the UPS for another seven days. No abnormal power fluctuations were recorded on either test and the workstation still dropped off the network three times.

The only thing left that I can think of is the operating system. I have seen one instance where changing the OS solved the problem. I admit this is a long shot but that's all that's left. We'll be changing the OS on the 26th and will post the results a few days later...

--

Dave
-----------------------------------------------------------------
I wonder why evolution hasn't got around to creating an undo button ....



"w_tom" <(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
> Assuming a hardware problems:
> 1) Route the NIC cable by a completely different path, such as
> a temporary connection down the hall. Then see what happens.
>
> 2) Run an extension cord from same wall receptacle that the
> 'faulty' station used to a wall receptacle at the hub. Then
> using a 3.5 digital volt meter, check voltages between that
> 'station' wall receptacle and voltages on the 'hub' wall
> receptacle. For example, no difference greater than 2 volts
> AC between each safety ground and both receptacle's neutral
> wires (4 measurements). Less than 10 volts difference between
> neutrals for each receptacle. IOW extension cord permits
> measuring these voltages from different parts of the building
> while everything else in that building is in operation.
>
> 3) Network cards have extensive diagnostics. Have seen where
> every diagnostic works just fine, but then the NIC fails only
> when doing long data exchanges with both NICs. Diagnostics
> require that both NICs be from same manufacturer. Both NICs
> can exchange test data in comprehensive testing without
> interfering with the rest of the network. This test can
> locate strange intermittents exactly at the time that some
> external event causes the failure. To see exactly when
> failure happens is also a powerful tool.
>
> 4) How is shield connected? Shield should be connected to
> 'chassis' ground only at one one - not at both.
>
> 5) This is also where an oscilloscope comes in handy. In one
> case, saw the offending NIC cable that created rounded off
> waves. Once the loose connection was solved, the NIC output
> sharper (squarer) waves.
>
> 6) You are fooling yourself if you think a plug-in UPS is
> doing anything more than battery backup. Plug-in UPSes
> connect computers directly to AC mains when not in battery
> backup mode. They are careful to word things so that some
> people 'feel' they provide all kinds of additional
> protection. In reality, the only thing they do reliably is
> protect data from blackouts and brownouts. Many can output
> quite dirty (spikey) power when in battery backup mode. No
> problem since that dirty electricity is not a problem to any
> properly constructed computer.
>
> For example, this UPS in battery backup mode outputs two 200
> volt square waves with a 270 volt spike between those square
> waves. This is called 120 VAC. Yes. It is equivalent to 120
> VAC. And since computer power supplies are so resilient, then
> that very dirty backup power does not upset computers. But
> when getting others to promote the myth, that same UPS
> manufacturer would have you 'feel' is provides cleaner power.
> It is just another reason why that oscilloscope is such a
> powerful tool.
>
> You don't need the scope. Just appreciate that the UPS is
> not a solution to power problems that cause hardware
> problems. Suggestion #3 is the first one I would try. It
> would demonstrate whether this is a hardware or software
> problem.
>
> Dave wrote:
> > No problem - when I finally get it figured out I'll be sure to
> > post the solution that worked for me.
> >
> > Dave
> >>
> >>>-----Original Message-----
> >>> With over five years installing and maintaining networks
> >>> I thought I'd seen just about everything but this one has
> >>> me stumped and I'm running out of places to look for
> >>> solutions... Here's the short version...
> >>>
> >>> Stand-alone Win2K File and Print Server, workstations
> >>> running a mix of Win98SE, Win2K and XP Pro. All
> >>> workstations use dynamic IP from DHCP server.
> >>>
> >>> The server as well as all PCs, hubs, routers, and
> >>> switches are plugged into APC UPS units. All cabling is
> >>> surge protected at the hub, router or switch as applicable
> >>> using APC ProtectNet devices.
> >>>
> >>>The problem...
> >>>
> >>> One particular workstation is randomly losing the
> >>> connection to the LAN. It might happen two or three times
> >>> in one day, or it could go for several weeks without a
> >>> problem.
> >>>
> >>>Troubleshooting steps so far include...
> >>>....New NIC
> >>>....New patch cables
> >>>....Changed ports on the switch
> >>>....Gave the workstation a static IP address
> >>>....Deleted and recreated the user account on the server
> >>>....Replaced the problem PC with a different one - the
> >> problem did not move with the PC.
> >>>
> >>> Thinking it was a bad cable run, we replaced the entire
> >>> CAT5e cable run from the wall jack to the switch using
> >>> shielded cable - problem is still there.
> > >
> >>> Any and all help would be most appreciated.

Hello Doug

This location is pretty much standardized on the 3COM NICs. Even so, I have tried an Intel PILA8460BN PRO/100+ and a D-Link DFE-530TX+.

All PCs are non-integrated white box systems we build ourselves and all NICs are PCI add-in cards. The OS on this particular machine was initially Win98SE but we have swapped out the entire unit with a brand new Win2K Pro box for testing with the same results -

When we took over the site, the original hub was a NetGear but that has been replaced (for testing) with a 3COM OfficeConnect Dual Speed Switch (Used but known good - we have been using it in the shop w/o problem).

At this point, all hardware (NICs and hubs/switches) between the problem machine and the server are all known good 3COM products.

I'm thinking that it is a power problem like tom mentioned, but that it's being caused by someone plugging something in that shouldn't be... since it's the season for them, my money is on a space heater. But until (if) I can actually see something like this all I can do is take them at their word when they promise me that nothing like that's happening...

--

Dave
-----------------------------------------------------------------
I wonder why evolution hasn't got around to creating an undo button ...



"Douglas Wyant [MSFT]" <(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
Dave,
I have a couple questions for you.
1. What NIC's have you used that are demonstrating this problem? (other than the 3c905c-txm?)
2. Are the NIC's you have tried add-in PCI or integrated (LOM)?
3. What type of PC's are these in?
4. What OS?
5. What make / model switch/hub are you plugging into?

I am aware of certain Dell Workstation models w/ integrated 3Com NIC's that when connected to NetGear or Linksys hubs/switches can potentially demonstrate the problem you've described.

Thanks,

Doug Wyant [MSFT]
Microsoft Corporation

--
Please do not send e-mail directly to this alias. This alias is for newsgroup purposes only.

This posting is provided "AS IS" with no warranties, and confers no rights.


"Dave" <dont-(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
Thanks for the reply and please accept my apologies for the delay in returning to this post.

With the exception of scoping the signals I think we've pretty much covered the possibilities suggested.

1) Did that - no change in symptoms
2) Outlet voltage is 121.3 VAC at the workstation; 121.2 at the hub; 121.0 at the router and 122 at the server.
3) NIC diags show no problem - all NICs in use are 3COM 3C-509C-TXM
4) Shields are grounded at one end only and all cabling meets or exceeds EIA/TIA 568B
5) We do have a quad trace scope available but after replacing the NIC in the offending workstation, and then replacing the entire workstation with a known good system, we don't think malformed signal is the cause. We also replaced the hub and router at the same time as the workstation replacement.
6) Good points all. I'm familiar with the limitations and inner workings of various UPS units having been a missile systems tech for several years in the military.

Grasping straws, I borrowed a Fluke 43B Power Quality Analyzer and left it connected to the outlet for seven days, then moved it to the output side of the UPS for another seven days. No abnormal power fluctuations were recorded on either test and the workstation still dropped off the network three times.

The only thing left that I can think of is the operating system. I have seen one instance where changing the OS solved the problem. I admit this is a long shot but that's all that's left. We'll be changing the OS on the 26th and will post the results a few days later...

--

Dave
-----------------------------------------------------------------
I wonder why evolution hasn't got around to creating an undo button ....



"w_tom" <(E-Mail Removed)> wrote in message news:(E-Mail Removed)...
> Assuming a hardware problems:
> 1) Route the NIC cable by a completely different path, such as
> a temporary connection down the hall. Then see what happens.
>
> 2) Run an extension cord from same wall receptacle that the
> 'faulty' station used to a wall receptacle at the hub. Then
> using a 3.5 digital volt meter, check voltages between that
> 'station' wall receptacle and voltages on the 'hub' wall
> receptacle. For example, no difference greater than 2 volts
> AC between each safety ground and both receptacle's neutral
> wires (4 measurements). Less than 10 volts difference between
> neutrals for each receptacle. IOW extension cord permits
> measuring these voltages from different parts of the building
> while everything else in that building is in operation.
>
> 3) Network cards have extensive diagnostics. Have seen where
> every diagnostic works just fine, but then the NIC fails only
> when doing long data exchanges with both NICs. Diagnostics
> require that both NICs be from same manufacturer. Both NICs
> can exchange test data in comprehensive testing without
> interfering with the rest of the network. This test can
> locate strange intermittents exactly at the time that some
> external event causes the failure. To see exactly when
> failure happens is also a powerful tool.
>
> 4) How is shield connected? Shield should be connected to
> 'chassis' ground only at one one - not at both.
>
> 5) This is also where an oscilloscope comes in handy. In one
> case, saw the offending NIC cable that created rounded off
> waves. Once the loose connection was solved, the NIC output
> sharper (squarer) waves.
>
> 6) You are fooling yourself if you think a plug-in UPS is
> doing anything more than battery backup. Plug-in UPSes
> connect computers directly to AC mains when not in battery
> backup mode. They are careful to word things so that some
> people 'feel' they provide all kinds of additional
> protection. In reality, the only thing they do reliably is
> protect data from blackouts and brownouts. Many can output
> quite dirty (spikey) power when in battery backup mode. No
> problem since that dirty electricity is not a problem to any
> properly constructed computer.
>
> For example, this UPS in battery backup mode outputs two 200
> volt square waves with a 270 volt spike between those square
> waves. This is called 120 VAC. Yes. It is equivalent to 120
> VAC. And since computer power supplies are so resilient, then
> that very dirty backup power does not upset computers. But
> when getting others to promote the myth, that same UPS
> manufacturer would have you 'feel' is provides cleaner power.
> It is just another reason why that oscilloscope is such a
> powerful tool.
>
> You don't need the scope. Just appreciate that the UPS is
> not a solution to power problems that cause hardware
> problems. Suggestion #3 is the first one I would try. It
> would demonstrate whether this is a hardware or software
> problem.
>
> Dave wrote:
> > No problem - when I finally get it figured out I'll be sure to
> > post the solution that worked for me.
> >
> > Dave
> >>
> >>>-----Original Message-----
> >>> With over five years installing and maintaining networks
> >>> I thought I'd seen just about everything but this one has
> >>> me stumped and I'm running out of places to look for
> >>> solutions... Here's the short version...
> >>>
> >>> Stand-alone Win2K File and Print Server, workstations
> >>> running a mix of Win98SE, Win2K and XP Pro. All
> >>> workstations use dynamic IP from DHCP server.
> >>>
> >>> The server as well as all PCs, hubs, routers, and
> >>> switches are plugged into APC UPS units. All cabling is
> >>> surge protected at the hub, router or switch as applicable
> >>> using APC ProtectNet devices.
> >>>
> >>>The problem...
> >>>
> >>> One particular workstation is randomly losing the
> >>> connection to the LAN. It might happen two or three times
> >>> in one day, or it could go for several weeks without a
> >>> problem.
> >>>
> >>>Troubleshooting steps so far include...
> >>>....New NIC
> >>>....New patch cables
> >>>....Changed ports on the switch
> >>>....Gave the workstation a static IP address
> >>>....Deleted and recreated the user account on the server
> >>>....Replaced the problem PC with a different one - the
> >> problem did not move with the PC.
> >>>
> >>> Thinking it was a bad cable run, we replaced the entire
> >>> CAT5e cable run from the wall jack to the switch using
> >>> shielded cable - problem is still there.
> > >
> >>> Any and all help would be most appreciated.