[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Wed, 5 Apr 2017 14:25:17 -0600
schrieb the...@sys-concept.com:

> On 04/05/2017 12:22 PM, Kai Krakow wrote:
> > Am Wed, 5 Apr 2017 09:40:32 -0600
> > schrieb the...@sys-concept.com:
> >   
> [snip]
> >>
> >> The problem with the cable was incorrect cross-over, the cable was
> >> installed by a contractor few years back and terminated by them.
> >>
> >> When I initially tested the cable with cable tester one end was
> >> connected to a switch I connected the tester to another and all the
> >> lights were working in order.  Wrong assumption.
> >> The switch will respond to any ping and give a false readout.  
> > 
> > May I suggest that you use the active part of the tester always on
> > the side where the switch would be connected? This sorts out such
> > problems. ;-)  
> 
> I think this cable tester was only designed to work with passive
> receiver (that came with it) on the other end.
> Using active part with switch doesn't give meaningful reading, it only
> confirm that that are no broken wires inside.

No it does not. Switches and other devices may act upon such signals in
any way possible, e.g. being broken after attaching such a tester.

That is one reason why I recommended to use the active part of the
tester on the side where the switch would be connected: It forces you
to unplug the cable from the switch and plug it into the tester.


-- 
Regards,
Kai

Replies to list-only preferred.

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[thelma]

On 04/05/2017 06:35 PM, Daniel Frey wrote:
> On 04/05/2017 01:25 PM, the...@sys-concept.com wrote:
>>
>> I think this cable tester was only designed to work with passive
>> receiver (that came with it) on the other end.
>> Using active part with switch doesn't give meaningful reading, it only
>> confirm that that are no broken wires inside.
>>
> 
> It is possible to damage a switch port by hooking one of the cheap
> testers directly to one of the switch ports, just a warning. The testers
> that actually do speed certification testing are safe. If the tester you
> have only has LEDs to indicate continuity don't hook it up to a switch port!
> 
> Dan

Thank you for the warning.
I didn't know that.

--
Thelma

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Daniel Frey]

On 04/05/2017 01:25 PM, the...@sys-concept.com wrote:
> 
> I think this cable tester was only designed to work with passive
> receiver (that came with it) on the other end.
> Using active part with switch doesn't give meaningful reading, it only
> confirm that that are no broken wires inside.
> 

It is possible to damage a switch port by hooking one of the cheap
testers directly to one of the switch ports, just a warning. The testers
that actually do speed certification testing are safe. If the tester you
have only has LEDs to indicate continuity don't hook it up to a switch port!

Dan

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[thelma]

On 04/05/2017 12:22 PM, Kai Krakow wrote:
> Am Wed, 5 Apr 2017 09:40:32 -0600
> schrieb the...@sys-concept.com:
> 
[snip]
>>
>> The problem with the cable was incorrect cross-over, the cable was
>> installed by a contractor few years back and terminated by them.
>>
>> When I initially tested the cable with cable tester one end was
>> connected to a switch I connected the tester to another and all the
>> lights were working in order.  Wrong assumption.
>> The switch will respond to any ping and give a false readout.
> 
> May I suggest that you use the active part of the tester always on the
> side where the switch would be connected? This sorts out such
> problems. ;-)

I think this cable tester was only designed to work with passive
receiver (that came with it) on the other end.
Using active part with switch doesn't give meaningful reading, it only
confirm that that are no broken wires inside.

> 
>> I suppose to connect one end to transmitter and the other end to the
>> receiver (that came with it) and I've noticed incorrect order.
>> From Left:
>> Orange/White
>> Orange
>> Blue/Stripe
>> Blue
>> Green/Stripe
>> Green
>> Brown/Strip
>> Brown
>>
>> Re-termination fixed the problem.
> 
> Good that you found it. Does it work correctly now?

It was late last night, so I didn't have a chance to test it, but will
do it this week.

--
Thelma

[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Wed, 5 Apr 2017 09:40:32 -0600
schrieb the...@sys-concept.com:

> On 04/05/2017 08:14 AM, Daniel Frey wrote:
> > On 04/04/2017 03:11 PM, Kai Krakow wrote:  
> >>
> >> That's a good note.
> >>
> >> I think punch down is similar to LSA plus here in Germany, tho the
> >> tools look very different. But with LSA you also push the wire
> >> between to cutting contacts.
> >>  
> > 
> > Yeah, it sounds like it works the same here: a wire is pushed
> > through a V-bladed contact via a tool.
> > 
> > Dan  
> 
> The problem with the cable was incorrect cross-over, the cable was
> installed by a contractor few years back and terminated by them.
> 
> When I initially tested the cable with cable tester one end was
> connected to a switch I connected the tester to another and all the
> lights were working in order.  Wrong assumption.
> The switch will respond to any ping and give a false readout.

May I suggest that you use the active part of the tester always on the
side where the switch would be connected? This sorts out such
problems. ;-)

> I suppose to connect one end to transmitter and the other end to the
> receiver (that came with it) and I've noticed incorrect order.
> From Left:
> Orange/White
> Orange
> Blue/Stripe
> Blue
> Green/Stripe
> Green
> Brown/Strip
> Brown
> 
> Re-termination fixed the problem.

Good that you found it. Does it work correctly now?


-- 
Regards,
Kai

Replies to list-only preferred.

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[thelma]

On 04/05/2017 08:14 AM, Daniel Frey wrote:
> On 04/04/2017 03:11 PM, Kai Krakow wrote:
>>
>> That's a good note.
>>
>> I think punch down is similar to LSA plus here in Germany, tho the
>> tools look very different. But with LSA you also push the wire between
>> to cutting contacts.
>>
> 
> Yeah, it sounds like it works the same here: a wire is pushed through a
> V-bladed contact via a tool.
> 
> Dan

The problem with the cable was incorrect cross-over, the cable was
installed by a contractor few years back and terminated by them.

When I initially tested the cable with cable tester one end was
connected to a switch I connected the tester to another and all the
lights were working in order.  Wrong assumption.
The switch will respond to any ping and give a false readout.

I suppose to connect one end to transmitter and the other end to the
receiver (that came with it) and I've noticed incorrect order.
>From Left:
Orange/White
Orange
Blue/Stripe
Blue
Green/Stripe
Green
Brown/Strip
Brown

Re-termination fixed the problem.

--
Thelma

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Daniel Frey]

On 04/04/2017 03:11 PM, Kai Krakow wrote:
> 
> That's a good note.
> 
> I think punch down is similar to LSA plus here in Germany, tho the
> tools look very different. But with LSA you also push the wire between
> to cutting contacts.
> 

Yeah, it sounds like it works the same here: a wire is pushed through a
V-bladed contact via a tool.

Dan

[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Tue, 4 Apr 2017 15:05:57 -0700
schrieb Daniel Frey :

> On 04/04/2017 02:49 PM, Kai Krakow wrote:
> > No, if you have the same wrong on both sides, the LEDs will still
> > show correct blinking order. Think of it like this: If you use order
> > 7-5-3-1-2-4-6-8 on both sides, blinking LED 1 on one side will blink
> > the same LED on the other side because they both connect to wire 7.
> > But the twisted pairs that should be twisted are no longer because
> > now you connected pair 1 in the connector to wire 7 and 5 which
> > belong to different pairs in the wire. The wire twists pair (7,8)
> > and (4,5). On the long run, interference now cannot be canceled out
> > because this only works if wires are twisted in the same pair. The
> > connector (and ethernet standard) expects the following pairs on
> > the connector:
> > 
> > A-A-B-C-C-B-D-D
> > 1-2-3-4-5-6-7-8
> > 
> > For electrical reasons it also expects a white wire alternating
> > with a colored pair, which makes the following pairs:
> > 
> > A = (1,2)
> > B = (3,6)
> > C = (5,4)
> > D = (7,8)
> > 
> > Or:
> > 
> > A-a B c-C b D-d
> > ^_^
> > 
> > With the capital letters being either all white or all colored (this
> > doesn't depend as long as it's the same on both sides).
> > 
> > You could open the problematic wall outlets and check the cabling
> > yourself. Keep in mind that unmounting the wall outlet may make the
> > problem of bent cables even worse due to moving and bending the
> > wires even more.   
> 
> I'll pitch in one interesting thing of note: the local supplier here
> went with el-cheapo Chinese made punch down and crimp ends some years
> ago. These are supposed to be 22 AWG wire size, but after dealing with
> bad connections I found out these Chinese punch downs were more like
> 24-26 AWG, meaning when punched down it would not make proper contact
> with the properly spec'ed wire.

That's a good note.

I think punch down is similar to LSA plus here in Germany, tho the
tools look very different. But with LSA you also push the wire between
to cutting contacts.

-- 
Regards,
Kai

Replies to list-only preferred.

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Daniel Frey]

On 04/04/2017 02:49 PM, Kai Krakow wrote:
> No, if you have the same wrong on both sides, the LEDs will still show
> correct blinking order. Think of it like this: If you use order
> 7-5-3-1-2-4-6-8 on both sides, blinking LED 1 on one side will blink
> the same LED on the other side because they both connect to wire 7. But
> the twisted pairs that should be twisted are no longer because now you
> connected pair 1 in the connector to wire 7 and 5 which belong to
> different pairs in the wire. The wire twists pair (7,8) and (4,5). On
> the long run, interference now cannot be canceled out because this
> only works if wires are twisted in the same pair. The connector (and
> ethernet standard) expects the following pairs on the connector:
> 
> A-A-B-C-C-B-D-D
> 1-2-3-4-5-6-7-8
> 
> For electrical reasons it also expects a white wire alternating with a
> colored pair, which makes the following pairs:
> 
> A = (1,2)
> B = (3,6)
> C = (5,4)
> D = (7,8)
> 
> Or:
> 
> A-a B c-C b D-d
> ^_^
> 
> With the capital letters being either all white or all colored (this
> doesn't depend as long as it's the same on both sides).
> 
> You could open the problematic wall outlets and check the cabling
> yourself. Keep in mind that unmounting the wall outlet may make the
> problem of bent cables even worse due to moving and bending the wires
> even more. 

I'll pitch in one interesting thing of note: the local supplier here
went with el-cheapo Chinese made punch down and crimp ends some years
ago. These are supposed to be 22 AWG wire size, but after dealing with
bad connections I found out these Chinese punch downs were more like
24-26 AWG, meaning when punched down it would not make proper contact
with the properly spec'ed wire.

Dan

[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Tue, 4 Apr 2017 15:05:13 -0600
schrieb the...@sys-concept.com:

> On 04/04/2017 02:56 PM, Kai Krakow wrote:
> > Am Tue, 4 Apr 2017 14:28:23 -0600
> > schrieb the...@sys-concept.com:
> >   
> [snip]
> >>
> >> I have reconnected another cable and the unit in remote location
> >> works. Both cable have a good pinout but one is working and the
> >> other is not. Both cable are sunning inside wall (I presume same
> >> path). Without special tools/testing equipment it is hard to trace
> >> these problems.  
> > 
> > You could try the problematic cable with only 100 MBit. If this
> > works, I'm pretty sure that some of the wires are broken or have
> > incorrect order. Keep in mind, tho, that the inverse assumption of
> > such tests is not true.  
> 
> Yes, the testing took was cheap it came with the stripper.
> Though, if the cable order was wrong, wouldn't the light on the tester
> jump in different order?  The light on the tester lights up
> sequentially, so I assume the order is correct.  Besides that "bad"
> cable was working OK for a day.

No, if you have the same wrong on both sides, the LEDs will still show
correct blinking order. Think of it like this: If you use order
7-5-3-1-2-4-6-8 on both sides, blinking LED 1 on one side will blink
the same LED on the other side because they both connect to wire 7. But
the twisted pairs that should be twisted are no longer because now you
connected pair 1 in the connector to wire 7 and 5 which belong to
different pairs in the wire. The wire twists pair (7,8) and (4,5). On
the long run, interference now cannot be canceled out because this
only works if wires are twisted in the same pair. The connector (and
ethernet standard) expects the following pairs on the connector:

A-A-B-C-C-B-D-D
1-2-3-4-5-6-7-8

For electrical reasons it also expects a white wire alternating with a
colored pair, which makes the following pairs:

A = (1,2)
B = (3,6)
C = (5,4)
D = (7,8)

Or:

A-a B c-C b D-d
^_^

With the capital letters being either all white or all colored (this
doesn't depend as long as it's the same on both sides).

You could open the problematic wall outlets and check the cabling
yourself. Keep in mind that unmounting the wall outlet may make the
problem of bent cables even worse due to moving and bending the wires
even more. Usually, the company that installed the cabling should've
done a frequency spectrum test on each wire pair. If you didn't get it
you should ask for it. In my company, we deny any network problems of
our delivered equipment unless such a test has been done and presented
to us. Such a test may cost a few extra bucks but should be part of
such an installation (done by the electrician before handling the
project over to its customer).

Depending on the inside of the RJ45 wall outlet, you can fix it
yourself. For the cheaper LSA based internal connectors you need an LSA
tool (it's inexpensive). Don't try to use a screw driver to mount the
wires. The more expensive modular connectors are easy to mount: Just
insert the wire pairs properly into the holes, properly and cleanly cut
the wires off the other end, and push the connector module in place
(take care of proper alignment). The modular connectors also
properly connect shielding, so properly connect that. Usually those
connectors come with a small manual how to do it. Order new ones, don't
reuse. Cut off 1-2 inches from the old connector cabling.

At least in Germany I know those two kinds of connector outlets.

> And yes, the room the cable is passing by has all kind or x-ray
> machine.

There's definitely problems with x-ray machines. Usually, those are
connected by fiber optics (at least dental x-ray machines). You want to
find a separate cabling way for your network, and also ensure that the
power circuits are different between those machines and your network
equipment.

Ensure that shielding is properly passed along the complete cabling,
including the patch cable from the wall outlet to your machine. Proper
shielding is essential in such environments.

We had one issue at a company running cooling generators and having
unstable network. It was eventually resolved somehow but we think that
the generators induced some non-harmonic distortion into the cabling of
the complete building. I think it was resolved when they rechecked
proper shielding on all cabling and machines.

> I'll try to test it 100Mbit (limit the speed); just need to find out
> how.

Maybe this can be done with ethtool.

-- 
Regards,
Kai

Replies to list-only preferred.

Re: [gentoo-user] Re: soliciting a DHCP lease / carrier lost

[thelma]

On 04/04/2017 02:56 PM, Kai Krakow wrote:
> Am Tue, 4 Apr 2017 14:28:23 -0600
> schrieb the...@sys-concept.com:
> 
[snip]
>>
>> I have reconnected another cable and the unit in remote location
>> works. Both cable have a good pinout but one is working and the other
>> is not. Both cable are sunning inside wall (I presume same path).
>> Without special tools/testing equipment it is hard to trace these
>> problems.
> 
> You could try the problematic cable with only 100 MBit. If this works,
> I'm pretty sure that some of the wires are broken or have incorrect
> order. Keep in mind, tho, that the inverse assumption of such tests is
> not true.

Yes, the testing took was cheap it came with the stripper.
Though, if the cable order was wrong, wouldn't the light on the tester
jump in different order?  The light on the tester lights up
sequentially, so I assume the order is correct.  Besides that "bad"
cable was working OK for a day.
And yes, the room the cable is passing by has all kind or x-ray machine.

I'll try to test it 100Mbit (limit the speed); just need to find out how.

--
Thelma.

[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Tue, 4 Apr 2017 14:28:23 -0600
schrieb the...@sys-concept.com:

> On 04/04/2017 10:02 AM, Mick wrote:
> > On Tuesday 04 Apr 2017 09:12:16 the...@sys-concept.com wrote:  
> >> On 04/04/2017 01:26 AM, Mick wrote:  
>  [...]  
> >   
>  [...]  
> > 
> > This may merely indicate they have been wired correctly (pin to
> > pin).  Unless your tester is 'intelligent' to also measure things
> > like attenuation, DC loop resistance and cross talk and it can also
> > calculate attenuation to cross talk ratio, you cannot be sure your
> > cable will perform to specification.
> > 
> >   
>  [...]  
> >>
> >> Shouldn't CAT5 be able to handle 100m run?
> >> Am not sure I understand, "keep their runs separate from mains
> >> cables"?  
> > 
> > Cat5e should be able to perform as specified in lengths up to 100m,
> > when correctly terminated and without high cross talk.  If your
> > ethernet cable installation is running parallel to mains power and
> > in close physical proximity, it may pick up noise, which will
> > reduce its performance.  It is better where ethernet and mains runs
> > come together to cross them at 90 degrees angles to minimise the
> > effect of interference.
> > 
> > Either way, you have lost carrier errors.  Random google result on
> > causes of lost carrier errors, in case it helps:
> > 
> > https://supportforums.cisco.com/discussion/9543606/what-causes-output-errors-ethernet-interface
> >   
> 
> I have reconnected another cable and the unit in remote location
> works. Both cable have a good pinout but one is working and the other
> is not. Both cable are sunning inside wall (I presume same path).
> Without special tools/testing equipment it is hard to trace these
> problems.

You could try the problematic cable with only 100 MBit. If this works,
I'm pretty sure that some of the wires are broken or have incorrect
order. Keep in mind, tho, that the inverse assumption of such tests is
not true.

-- 
Regards,
Kai

Replies to list-only preferred.

[gentoo-user] Re: soliciting a DHCP lease / carrier lost

[Kai Krakow]

Am Tue, 4 Apr 2017 09:12:16 -0600
schrieb the...@sys-concept.com:

> On 04/04/2017 01:26 AM, Mick wrote:
> > On Monday 03 Apr 2017 20:21:28 the...@sys-concept.com wrote:  
> >> The new box I installed in remote location has a problem obtaining
> >> IP address.  The box was working perfectly on my local LAN.
> >>
> >> In remote location I assigned static IP to it 10.10.0.5  
> > 
> > Where and how?  At the router IP address table, against the PC's
> > MAC address? At the PC itself using a static IP address and gateway
> > in /etc/conf.d/net?  
> 
> The remote location router runs DD-WRT (dhcpd), so all static IP's
> are assigned via DD-WRT and MAC address.
> 
> >> Previously this IP was assigned to a Virtual Box but I no longer
> >> use it, so I assign this IP to a new box.
> >>
> >> The box was working for a day, but now when I boot the box I get
> >> - soliciting a DHCP lease
> >> - carrier lost  
> > 
> > The "carrier lost" error indicates a link going down.  The lease
> > renewal is likely to fail at least while the link is down.  The
> > link failure may be due to an electrical hard fault, e.g. faulty
> > Cat5e cable, RJ45 socket/plug; or due to high electromagnetic
> > interference.  Check the router stats for carrier lost errors.  If
> > the counters show the link is being dropped regularly you should
> > try to eliminate each component in the circuit the cause of the
> > fault.  
> 
> The cable tester I have (cheap) is showing the CAT5 cable is OK.
> http://www.primecables.com/p-309139-cab-ss35407-tester-network-cable-tester-crimping-tools-combo-for-rj-45-rj-11-primecables?gclid=CJzU2JKJi9MCFQYMaQodEyIHeA

Such testers just test the correct order of the wires and that at
least some signal reaches the other end. They don't test signal
quality. If you use cheap connectors (the ones made only of plastic),
the shielding won't be connected from one connector to the other which
may be part of your problem. Also, such testers don't detect if you
bent your cable too much: A bent wire still shows perfect connectivity
for simple LED testers but high frequency signals get distorted too
much by such cable issues.

Also, they don't test the pairs are correctly twisted which each other.
You could simply use wrong pairing like (1,2)(3,4)(5,6)(7,8) and the
tester will show every wire okay. But the signals won't run correctly
because twisted pair cables use pairs of (1,2)(3,6)(5,4)(7,8) - it is
even important that one pair is reversed as you can see. In the
connector it always has to be color-white-color-white-color-white
alternating where one pair is split around the middle pair. Only the
correct ordering of wires will ensure proper signal quality.

If you made your cables yourself, you should check that.

Here's a very good reference and explanation I always use:
http://www.elektronik-kompendium.de/sites/net/0510151.htm

It's German but the tables should be easy enough to understand. For the
text you can try some translation tool.

 
> >> Could the old IP get stuck somewhere in DD-WRT router?
> >>
> >> ping 10.10.0.5 - gives me no response.  
> > 
> > You may find arping a better instrument for investigating the use
> > of IP addresses in your LAN.
> > 
> >   
> >> The Cat5 is about 15-20meter long, I test it with a cable tester,
> >> it is good (all the lights light up in correct order).
> >> Cable is plugged in into a new switch.  
> > 
> > Long cables are more susceptible to electromagnetic interference -
> > keep their runs separate from mains cables.  
> 
> Shouldn't CAT5 be able to handle 100m run?
> Am not sure I understand, "keep their runs separate from mains
> cables"?
> 
> Electromagnetic interference - could be a problem, and it is hard to
> troubleshoot. 

Wasn't it you that asked before about IP address assignment problems
which appear in one location but not in the other?

Maybe it is simply an environmental problem: Your cables may run along
main power cables or high power cables attached to power demanding
devices. This may well impose problems.

If this is the problem you should consider using fiber optics on the
longer runs or use CAT7 cabling. Tho, you can usually not attach normal
crimping RJ45 connectors to such cables as the cables are too thick.
There are special connectors for such cables. Keep also in mind that
real CAT7 connectors are incompatible to your RJ45 switches. You would
use normal CAT6 connectors which more or less caps your cabling at CAT6
then but still you benefit from CAT7 cabling on the long runs.


-- 
Regards,
Kai

Replies to list-only preferred.