Re: [pfSense] Slow speed on 100Base TX full duplex.

[Muhammad Yousuf Khan]

Thanks Alot guys, this is something can do with thorough testings but my
client does not want to work on this colo anymore and i have no more access
of these servers :(.

Thanks,
Yousuf

On Wed, Jan 13, 2016 at 9:44 AM, Ugo Bellavance  wrote:

> On 16-01-11 01:23 AM, Muhammad Yousuf Khan wrote:
>
>> I am remotely supporting one of my client who is using pfsense.  i have
>> been using pfsense for years and never face such issue in this
>> experience, the Client Co-location is recommending to use 100BaseTX full
>> duplex setting instead of Auto. i do not know why they required that since
>> i am not in US i never observe this settings recommended by colo people in
>> my country.
>>
>> - iperf speed test for LAN, between is 50Mbps  up and down
>> - but iperf test on WAN showing 10Mbps down and 5Mbps up.
>> - however my client is saying that assigned speed from colo is 100Mbps.
>>
>> now i can not find where is the issue. i suspect that issue is with
>> 100BaseTX setting.
>>
>
> You generally have to configure your equipment the way the colo people
> tell you.  Have you communicated with the colo directly to get help? Also
> did you try just connecting a Windows or Linux laptop on the colo switch to
> see what kind of speed you get.  Are you sure the other endpoint of your
> iperf test can send and receive 100 mbps?
>
> Ugo
>
>
> ___
> pfSense mailing list
> https://lists.pfsense.org/mailman/listinfo/list
> Support the project with Gold! https://pfsense.org/gold
>
___
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Ugo Bellavance]

On 16-01-11 01:23 AM, Muhammad Yousuf Khan wrote:

I am remotely supporting one of my client who is using pfsense.  i have
been using pfsense for years and never face such issue in this
experience, the Client Co-location is recommending to use 100BaseTX full
duplex setting instead of Auto. i do not know why they required that since
i am not in US i never observe this settings recommended by colo people in
my country.

- iperf speed test for LAN, between is 50Mbps  up and down
- but iperf test on WAN showing 10Mbps down and 5Mbps up.
- however my client is saying that assigned speed from colo is 100Mbps.

now i can not find where is the issue. i suspect that issue is with
100BaseTX setting.


You generally have to configure your equipment the way the colo people 
tell you.  Have you communicated with the colo directly to get help? 
Also did you try just connecting a Windows or Linux laptop on the colo 
switch to see what kind of speed you get.  Are you sure the other 
endpoint of your iperf test can send and receive 100 mbps?


Ugo

___
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Muhammad Yousuf Khan]

em0@pci0:4:0:0: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
hdr=0x00
class  = network
subclass   = ethernet
em1@pci0:4:0:1: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
hdr=0x00
class  = network
subclass   = ethernet

We had a switch in b/w Pfsense and Colo uplink. we even removed that switch
and directly plug the cable with pfsense interface. but still getting the
same low bandwidth.

is it a good idea. to install two new interfaces of 100Mbps and set them to
Auto instead of making it static 100Base TX full dublex out of Gig
Interfaces.  ?

Any help will be highly appreciated.

Thanks,
Yousuf


On Mon, Jan 11, 2016 at 6:03 PM, C. R. Oldham  wrote:

> Re: pkg_add, try just 'pkg install' instead.
>
> Like Juan said, did you get them to try a different cable? Those errors
> are indicative of a bad Ethernet cable.
>
> Also, if the Ethernet chipset is a Realtek, there is a bug in the FreeBSD
> driver that affects auto negotiation with some switch hardware.
>
> --cro
>
>
> > On Jan 11, 2016, at 05:40, Muhammad Yousuf Khan 
> wrote:
> >
> > Here you go, yes there are error in the interfaces  i can not get more
> > detail as i can not run the command pkg_add  it is saying that command
> not
> > found however i know its a server board and it has two bultin LAN. 1 i am
> > using for WAN and For LAN.
> > here is CPU details.
> >
> > Intel(R) Xeon(R) CPU E5440 @ 2.83GHz
> > 8 CPUs: 2 package(s) x 4 core(s)
> >
> > Any guide will be highly appreciated.
> >
> > WAN interface (wan, em0)StatusupMAC addressxIPv4
> address
> > xxSubnet masGateway
> IPv4xxIPv6
> > Link Locaxxx
> >
> > ISP DNS serversxxxMTU1500Media100baseTX
> In/out
> > packets3709795/2014620 (3.06 GB/551.84 MB)In/out packets
> (pass)3709795/2014620
> > (3.06 GB/551.84 MB)In/out packets (block)90881/1 (6.59 MB/52 bytes)In/out
> > errors665/0Collisions0LAN interface (lan, em1)StatusupMAC address
> > xxIPv4 addressSubnet mask IPv4
> > IPv6 Link
> > LocalxxxMTU1500Media100baseTX
> > In/out packets1071425/2719703 (439.25 MB/2.78 GB)In/out
> > packets (pass)1071425/2719703 (439.25 MB/2.78 GB)In/out packets
> (block)2040/0
> > (174 KB/0 bytes)In/out errors2140/0Collisions0
> >
> > On Mon, Jan 11, 2016 at 2:16 PM, Juan Pablo 
> > wrote:
> >
> >> Hey, yes usually you should set 10/100/g to see when the link state
> >> changes, also if the auto protocol is not working or if the cable goes
> bad
> >> is easier to troubleshoot, have seen this on co-los worldwide.  in any
> >> case, setting 10/100 etc shouldnt affect the bandwidth. so the question
> >> here is: which Nic you are using? is it supported?
> >> do you see any network issue/crc issue, alert/errors, or something onthe
> >> logs? via the web interface check if there are any error on the
> interface
> >> counters.
> >> also: check with ifconfig 'interface name' for crc errors, and the
> >> advertised speeds, paste here the full output of the problematic
> interface.
> >>
> >>
> >> let us know how it goes.
> >>
> >>
> >> 2016-01-11 3:23 GMT-03:00 Muhammad Yousuf Khan :
> >>
> >>> I am remotely supporting one of my client who is using pfsense.  i have
> >>> been using pfsense for years and never face such issue in this
> >>> experience, the Client Co-location is recommending to use 100BaseTX
> full
> >>> duplex setting instead of Auto. i do not know why they required that
> >> since
> >>> i am not in US i never observe this settings recommended by colo people
> >> in
> >>> my country.
> >>>
> >>>
> >>>
> >>>  --Server 1
> >>>
> >>>/
> >>> colo switch>[WAN]pfsense[LAN]--+
> >>>
> >>>\
> >>>
> >>>  -Server 2
> >>>
> >>> - iperf speed test for LAN, between is 50Mbps  up and down
> >>> - but iperf test on WAN showing 10Mbps down and 5Mbps up.
> >>> - however my client is saying that assigned speed from colo is 100Mbps.
> >>>
> >>> now i can not find where is the issue. i suspect that issue is with
> >>> 100BaseTX setting.
> >>>
> >>> can anyone please guide me where i am doing wrong and what i can do to
> >> fix
> >>> this.
> >>> any help will be highly appreciated.
> >>>
> >>> Thanks,
> >>> Yousuf
> >>> ___
> >>> pfSense mailing list
> >>> https://lists.pfsense.org/mailman/listinfo/list
> >>> Support the project with Gold! https://pfsense.org/gold
> >> ___
> >> pfSense mailing list
> >> https://lists.pfsense.org/mailman/listinfo/list
> >> Support the project with Gold! https://pfsense.org/gold
> > ___
> > pfSense mailing list
> > https://lists.pfsense.org/mailman/listinfo/list
> > Support the project with Gold! 

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Seth Mos]

Op 11-1-2016 om 14:46 schreef Muhammad Yousuf Khan:
> em0@pci0:4:0:0: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
> class  = network
> subclass   = ethernet
> em1@pci0:4:0:1: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
> class  = network
> subclass   = ethernet
> 
> We had a switch in b/w Pfsense and Colo uplink. we even removed that switch
> and directly plug the cable with pfsense interface. but still getting the
> same low bandwidth.
> 
> is it a good idea. to install two new interfaces of 100Mbps and set them to
> Auto instead of making it static 100Base TX full dublex out of Gig
> Interfaces.  ?
> 
> Any help will be highly appreciated.

Only set the interface hard if the other side does that as well. You can
set it to 100 Mbit Full duplex, but if the other side does not force it
to the same value it will autonegotiate from the ISP or switch to Half
Duplex.

Overruns and runts galore.

If you put a unmanaged switch in between you will get this. If the ISP
switch is set to auto it will do the same thing.

So just leave it on auto, setting interfaces hard shouldn't be needed
anymore since we helft Nortel gear behind in the year 2000.

Cheers,
Seth
___
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold

Re: [pfSense] Slow speed on 100Base TX full duplex.

[C. R. Oldham]

First I would switch the interfaces to see if em1 delivers the same low 
performance as em0 when used as WAN.

If that doesn't work, try installing new interfaces.

You mentioned earlier that the colo told you to lock the WAN interface at 
100Mb/full-duplex. Our ISP was doing the same thing for a while. Autonegotiate 
used to be unreliable especially when 10Mbit interfaces were common.

Currently I am skeptical that the problem is on your end. Your error rate is 
not that high. Is your customer absolutely sure the colo didn't throttle their 
bandwidth?  Also is the colo sure they know which switch port goes to your 
customer? On my last go-around with our ISP they kept changing the settings on 
the wrong switch port on their end.

--cro

> On Jan 11, 2016, at 06:46, Muhammad Yousuf Khan  wrote:
> 
> em0@pci0:4:0:0: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
>class  = network
>subclass   = ethernet
> em1@pci0:4:0:1: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
>class  = network
>subclass   = ethernet
> 
> We had a switch in b/w Pfsense and Colo uplink. we even removed that switch
> and directly plug the cable with pfsense interface. but still getting the
> same low bandwidth.
> 
> is it a good idea. to install two new interfaces of 100Mbps and set them to
> Auto instead of making it static 100Base TX full dublex out of Gig
> Interfaces.  ?
> 
> Any help will be highly appreciated.
> 
> Thanks,
> Yousuf
> 
> 
>> On Mon, Jan 11, 2016 at 6:03 PM, C. R. Oldham  wrote:
>> 
>> Re: pkg_add, try just 'pkg install' instead.
>> 
>> Like Juan said, did you get them to try a different cable? Those errors
>> are indicative of a bad Ethernet cable.
>> 
>> Also, if the Ethernet chipset is a Realtek, there is a bug in the FreeBSD
>> driver that affects auto negotiation with some switch hardware.
>> 
>> --cro
>> 
>> 
>>> On Jan 11, 2016, at 05:40, Muhammad Yousuf Khan 
>> wrote:
>>> 
>>> Here you go, yes there are error in the interfaces  i can not get more
>>> detail as i can not run the command pkg_add  it is saying that command
>> not
>>> found however i know its a server board and it has two bultin LAN. 1 i am
>>> using for WAN and For LAN.
>>> here is CPU details.
>>> 
>>> Intel(R) Xeon(R) CPU E5440 @ 2.83GHz
>>> 8 CPUs: 2 package(s) x 4 core(s)
>>> 
>>> Any guide will be highly appreciated.
>>> 
>>> WAN interface (wan, em0)StatusupMAC addressxIPv4
>> address
>>> xxSubnet masGateway
>> IPv4xxIPv6
>>> Link Locaxxx
>>> 
>>> ISP DNS serversxxxMTU1500Media100baseTX
>> In/out
>>> packets3709795/2014620 (3.06 GB/551.84 MB)In/out packets
>> (pass)3709795/2014620
>>> (3.06 GB/551.84 MB)In/out packets (block)90881/1 (6.59 MB/52 bytes)In/out
>>> errors665/0Collisions0LAN interface (lan, em1)StatusupMAC address
>>> xxIPv4 addressSubnet mask IPv4
>>> IPv6 Link
>>> LocalxxxMTU1500Media100baseTX
>>> In/out packets1071425/2719703 (439.25 MB/2.78 GB)In/out
>>> packets (pass)1071425/2719703 (439.25 MB/2.78 GB)In/out packets
>> (block)2040/0
>>> (174 KB/0 bytes)In/out errors2140/0Collisions0
>>> 
>>> On Mon, Jan 11, 2016 at 2:16 PM, Juan Pablo 
>>> wrote:
>>> 
 Hey, yes usually you should set 10/100/g to see when the link state
 changes, also if the auto protocol is not working or if the cable goes
>> bad
 is easier to troubleshoot, have seen this on co-los worldwide.  in any
 case, setting 10/100 etc shouldnt affect the bandwidth. so the question
 here is: which Nic you are using? is it supported?
 do you see any network issue/crc issue, alert/errors, or something onthe
 logs? via the web interface check if there are any error on the
>> interface
 counters.
 also: check with ifconfig 'interface name' for crc errors, and the
 advertised speeds, paste here the full output of the problematic
>> interface.
 
 
 let us know how it goes.
 
 
 2016-01-11 3:23 GMT-03:00 Muhammad Yousuf Khan :
 
> I am remotely supporting one of my client who is using pfsense.  i have
> been using pfsense for years and never face such issue in this
> experience, the Client Co-location is recommending to use 100BaseTX
>> full
> duplex setting instead of Auto. i do not know why they required that
 since
> i am not in US i never observe this settings recommended by colo people
 in
> my country.
> 
> 
> 
> --Server 1
> 
>   /
> colo switch>[WAN]pfsense[LAN]--+
> 
>   \
> 
> -Server 2
> 
> - iperf speed test for LAN, between is 50Mbps  up and down
> - but iperf test on WAN showing 10Mbps down and 5Mbps up.
> - 

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Muhammad Yousuf Khan]

Here you go, yes there are error in the interfaces  i can not get more
detail as i can not run the command pkg_add  it is saying that command not
found however i know its a server board and it has two bultin LAN. 1 i am
using for WAN and For LAN.
here is CPU details.

Intel(R) Xeon(R) CPU E5440 @ 2.83GHz
8 CPUs: 2 package(s) x 4 core(s)

Any guide will be highly appreciated.

WAN interface (wan, em0)StatusupMAC addressxIPv4 address
xxSubnet masGateway IPv4xxIPv6
Link Locaxxx

ISP DNS serversxxxMTU1500Media100baseTX In/out
packets3709795/2014620 (3.06 GB/551.84 MB)In/out packets (pass)3709795/2014620
(3.06 GB/551.84 MB)In/out packets (block)90881/1 (6.59 MB/52 bytes)In/out
errors665/0Collisions0LAN interface (lan, em1)StatusupMAC address
xxIPv4 addressSubnet mask IPv4
IPv6 Link
LocalxxxMTU1500Media100baseTX
In/out packets1071425/2719703 (439.25 MB/2.78 GB)In/out
packets (pass)1071425/2719703 (439.25 MB/2.78 GB)In/out packets (block)2040/0
(174 KB/0 bytes)In/out errors2140/0Collisions0

On Mon, Jan 11, 2016 at 2:16 PM, Juan Pablo 
wrote:

> Hey, yes usually you should set 10/100/g to see when the link state
> changes, also if the auto protocol is not working or if the cable goes bad
> is easier to troubleshoot, have seen this on co-los worldwide.  in any
> case, setting 10/100 etc shouldnt affect the bandwidth. so the question
> here is: which Nic you are using? is it supported?
> do you see any network issue/crc issue, alert/errors, or something onthe
> logs? via the web interface check if there are any error on the interface
> counters.
> also: check with ifconfig 'interface name' for crc errors, and the
> advertised speeds, paste here the full output of the problematic interface.
>
>
> let us know how it goes.
>
>
> 2016-01-11 3:23 GMT-03:00 Muhammad Yousuf Khan :
>
> > I am remotely supporting one of my client who is using pfsense.  i have
> > been using pfsense for years and never face such issue in this
> > experience, the Client Co-location is recommending to use 100BaseTX full
> > duplex setting instead of Auto. i do not know why they required that
> since
> > i am not in US i never observe this settings recommended by colo people
> in
> > my country.
> >
> >
> >
> >   --Server 1
> >
> > /
> > colo switch>[WAN]pfsense[LAN]--+
> >
> > \
> >
> >   -Server 2
> >
> > - iperf speed test for LAN, between is 50Mbps  up and down
> > - but iperf test on WAN showing 10Mbps down and 5Mbps up.
> > - however my client is saying that assigned speed from colo is 100Mbps.
> >
> > now i can not find where is the issue. i suspect that issue is with
> > 100BaseTX setting.
> >
> > can anyone please guide me where i am doing wrong and what i can do to
> fix
> > this.
> > any help will be highly appreciated.
> >
> > Thanks,
> > Yousuf
> > ___
> > pfSense mailing list
> > https://lists.pfsense.org/mailman/listinfo/list
> > Support the project with Gold! https://pfsense.org/gold
> >
> ___
> pfSense mailing list
> https://lists.pfsense.org/mailman/listinfo/list
> Support the project with Gold! https://pfsense.org/gold
>
___
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Juan Pablo]

use Intel ethernets. even the cheap ones work as expected.
also ask them to set ports to 10 half 10 full 100 half, reset counters on
each change, you change to the same and see if in any works without errors.
or ask them to set it to auto. also change cables to something real.

its swimming with eyes closed with other than intel cards.
FYI:
https://doc.pfsense.org/index.php/Tuning_and_Troubleshooting_Network_Cards
https://www.pfsense.org/hardware/ at the bottom

sorry.


2016-01-11 10:46 GMT-03:00 Muhammad Yousuf Khan :

> em0@pci0:4:0:0: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
> class  = network
> subclass   = ethernet
> em1@pci0:4:0:1: class=0x02 card=0x15d9 chip=0x10968086 rev=0x01
> hdr=0x00
> class  = network
> subclass   = ethernet
>
> We had a switch in b/w Pfsense and Colo uplink. we even removed that switch
> and directly plug the cable with pfsense interface. but still getting the
> same low bandwidth.
>
> is it a good idea. to install two new interfaces of 100Mbps and set them to
> Auto instead of making it static 100Base TX full dublex out of Gig
> Interfaces.  ?
>
> Any help will be highly appreciated.
>
> Thanks,
> Yousuf
>
>
> On Mon, Jan 11, 2016 at 6:03 PM, C. R. Oldham  wrote:
>
> > Re: pkg_add, try just 'pkg install' instead.
> >
> > Like Juan said, did you get them to try a different cable? Those errors
> > are indicative of a bad Ethernet cable.
> >
> > Also, if the Ethernet chipset is a Realtek, there is a bug in the FreeBSD
> > driver that affects auto negotiation with some switch hardware.
> >
> > --cro
> >
> >
> > > On Jan 11, 2016, at 05:40, Muhammad Yousuf Khan 
> > wrote:
> > >
> > > Here you go, yes there are error in the interfaces  i can not get more
> > > detail as i can not run the command pkg_add  it is saying that command
> > not
> > > found however i know its a server board and it has two bultin LAN. 1 i
> am
> > > using for WAN and For LAN.
> > > here is CPU details.
> > >
> > > Intel(R) Xeon(R) CPU E5440 @ 2.83GHz
> > > 8 CPUs: 2 package(s) x 4 core(s)
> > >
> > > Any guide will be highly appreciated.
> > >
> > > WAN interface (wan, em0)StatusupMAC addressxIPv4
> > address
> > > xxSubnet masGateway
> > IPv4xxIPv6
> > > Link Locaxxx
> > >
> > > ISP DNS serversxxxMTU1500Media100baseTX
> > In/out
> > > packets3709795/2014620 (3.06 GB/551.84 MB)In/out packets
> > (pass)3709795/2014620
> > > (3.06 GB/551.84 MB)In/out packets (block)90881/1 (6.59 MB/52
> bytes)In/out
> > > errors665/0Collisions0LAN interface (lan, em1)StatusupMAC address
> > > xxIPv4 addressSubnet mask IPv4
> > > IPv6 Link
> > > LocalxxxMTU1500Media100baseTX
> > > In/out packets1071425/2719703 (439.25 MB/2.78 GB)In/out
> > > packets (pass)1071425/2719703 (439.25 MB/2.78 GB)In/out packets
> > (block)2040/0
> > > (174 KB/0 bytes)In/out errors2140/0Collisions0
> > >
> > > On Mon, Jan 11, 2016 at 2:16 PM, Juan Pablo  >
> > > wrote:
> > >
> > >> Hey, yes usually you should set 10/100/g to see when the link state
> > >> changes, also if the auto protocol is not working or if the cable goes
> > bad
> > >> is easier to troubleshoot, have seen this on co-los worldwide.  in any
> > >> case, setting 10/100 etc shouldnt affect the bandwidth. so the
> question
> > >> here is: which Nic you are using? is it supported?
> > >> do you see any network issue/crc issue, alert/errors, or something
> onthe
> > >> logs? via the web interface check if there are any error on the
> > interface
> > >> counters.
> > >> also: check with ifconfig 'interface name' for crc errors, and the
> > >> advertised speeds, paste here the full output of the problematic
> > interface.
> > >>
> > >>
> > >> let us know how it goes.
> > >>
> > >>
> > >> 2016-01-11 3:23 GMT-03:00 Muhammad Yousuf Khan :
> > >>
> > >>> I am remotely supporting one of my client who is using pfsense.  i
> have
> > >>> been using pfsense for years and never face such issue in this
> > >>> experience, the Client Co-location is recommending to use 100BaseTX
> > full
> > >>> duplex setting instead of Auto. i do not know why they required that
> > >> since
> > >>> i am not in US i never observe this settings recommended by colo
> people
> > >> in
> > >>> my country.
> > >>>
> > >>>
> > >>>
> > >>>  --Server 1
> > >>>
> > >>>/
> > >>> colo switch>[WAN]pfsense[LAN]--+
> > >>>
> > >>>\
> > >>>
> > >>>  -Server 2
> > >>>
> > >>> - iperf speed test for LAN, between is 50Mbps  up and down
> > >>> - but iperf test on WAN showing 10Mbps down and 5Mbps up.
> > >>> - however my client is saying that assigned speed from colo is
> 100Mbps.
> > >>>
> > >>> now i can not find where is the issue. 

Re: [pfSense] Slow speed on 100Base TX full duplex.

[Steve Yates]

Muhammad Yousuf Khan wrote on Mon, Jan 11 2016 at 12:23 am:

> - iperf speed test for LAN, between is 50Mbps  up and down
> - but iperf test on WAN showing 10Mbps down and 5Mbps up.
> - however my client is saying that assigned speed from colo is 100Mbps.

"full duplex" means the card sends and receives at the same time, so 
you normally want that on.

You said the colo port speed is 100 Mbps.  This is not necessarily the 
speed they have allowed for him, or the available bandwidth at the facility.  
If we imagine he is paying for a 50 Mbps connection the Ethernet port speed is 
still going to be 100 because the only choices are 10, 100, 1000, or 10 Gbit.

Likewise, if the colo has a lot of traffic, he may not get a 100 Mbps 
download speed when testing.

--

Steve Yates
ITS, Inc.



___
pfSense mailing list
https://lists.pfsense.org/mailman/listinfo/list
Support the project with Gold! https://pfsense.org/gold