CentOS - Sep 2008 - Need help with an odd issue I am experiencing

I have a server with dual nic's running centos 5.2 1 nic connects to the 
cable modem, which has 15m d and 2m up, but speed tests barely come back 
at 5mb down including road runner's internal speed test. I hook up my 
laptop to the same modem, same port, using the same cable and I get 17mb 
down. I have ruled out iptables, nics (because same result on both 
nics), cables, cable modem(with a brighthouse rep) any suggestions would 
be greatly appreciated. Thanks.

David Petruzzella Wrote:

-----Original Message-----
From: centos-bounces at centos.org [mailto:centos-bounces at centos.org] On
Behalf
Of David Petruzzella
Sent: Monday, September 08, 2008 10:09 PM
To: centos at centos.org
Subject: [CentOS] Need help with an odd issue I am experiencing

I have a server with dual nic's running centos 5.2 1 nic connects to the
cable modem, which has 15m d and 2m up, but speed tests barely come back at
5mb down including road runner's internal speed test. I hook up my laptop to
the same modem, same port, using the same cable and I get 17mb down. I have
ruled out iptables, nics (because same result on both nics), cables, cable
modem(with a brighthouse rep) any suggestions would be greatly appreciated.
Thanks.

**********************************************
JohnStanley Writes:

Maybe trying the vender specific drivers for the Nic Card in question on the
CentOS Server may solve the problem. Or a possibilty is if your Bonding or
teaming the nics there may be configuration issues in the way you have them
set up. Try just one Nic card by itself with out Bonding. You can use
"miitool" to set or force the transfer rate of the nic in question;
see "man
miitool" Since you do say you have dual nics try them both separtly like I
stated (Without The Bonding).

JohnStanley

> 
> I have a server with dual nic's running centos 5.2 1 nic connects to
the
> cable modem, which has 15m d and 2m up, but speed tests barely come back
> at 5mb down including road runner's internal speed test. I hook up my
> laptop to the same modem, same port, using the same cable and I get 17mb
> down. I have ruled out iptables, nics (because same result on both
> nics), cables, cable modem(with a brighthouse rep) any suggestions would
> be greatly appreciated. Thanks.
> 
David,

IME, it is typically (yet not always) a duplex and/or speed issue between
the NIC port and the switch or device it is talking too.

man ethtool

ethtool eth0

and check to see what speed and duplex are currently in use...

if duplex and/or speed are off I typically hard code a *managed* switch and
the centos box on whatever ports as necessary by adding this line or one
similar to it

ETHTOOL_OPTS="speed 100 duplex full autoneg off"

To this and other interface files as necessary

/etc/sysconfig/network-scripts/ifcfg-eth0

At this point a reboot will take care of it or if you don't want to reboot
now, Then I would make a little script called downup.sh

ifdown eth0
ifup eth0

and then I chmod 750 that script and then call it like this if from remote

./downup.sh &

So I don't lose connectivity...

There is probably a more elegant command like way to do that...

Be careful if remote box and YMMV

 - rh

Thanks I read about that, but I am having a hard time forcing duplex, I
don't remember the exact error it was giving me at the moment but from what
I read it looks like a driver issue.
---------- Original Message ----------------------------------
From: "RobertH" <roberth at abbacomm.net>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 07:30:30 -0700
>
>> 
>> I have a server with dual nic's running centos 5.2 1 nic connects
to the
>> cable modem, which has 15m d and 2m up, but speed tests barely come
back
>> at 5mb down including road runner's internal speed test. I hook up
my
>> laptop to the same modem, same port, using the same cable and I get
17mb
>> down. I have ruled out iptables, nics (because same result on both
>> nics), cables, cable modem(with a brighthouse rep) any suggestions
would
>> be greatly appreciated. Thanks.
>> 
>
>David,
>
>IME, it is typically (yet not always) a duplex and/or speed issue between
>the NIC port and the switch or device it is talking too.
>
>man ethtool
>
>ethtool eth0
>
>and check to see what speed and duplex are currently in use...
>
>if duplex and/or speed are off I typically hard code a *managed* switch and
>the centos box on whatever ports as necessary by adding this line or one
>similar to it
>
>ETHTOOL_OPTS="speed 100 duplex full autoneg off"
>
>To this and other interface files as necessary
>
>/etc/sysconfig/network-scripts/ifcfg-eth0
>
>At this point a reboot will take care of it or if you don't want to
reboot
>now, Then I would make a little script called downup.sh
>
>ifdown eth0
>ifup eth0
>
>and then I chmod 750 that script and then call it like this if from remote
>
>./downup.sh &
>
>So I don't lose connectivity...
>
>There is probably a more elegant command like way to do that...
>
>Be careful if remote box and YMMV
>
> - rh
>
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

I was thinkging drivers myself, I'm not bonding the nics btw.

---------- Original Message ----------------------------------
From: "John" <jses27 at gmail.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 01:34:32 -0400
> 
>
>
>David Petruzzella Wrote:
>
>-----Original Message-----
>From: centos-bounces at centos.org [mailto:centos-bounces at centos.org] On
Behalf
>Of David Petruzzella
>Sent: Monday, September 08, 2008 10:09 PM
>To: centos at centos.org
>Subject: [CentOS] Need help with an odd issue I am experiencing
>
>I have a server with dual nic's running centos 5.2 1 nic connects to the
>cable modem, which has 15m d and 2m up, but speed tests barely come back at
>5mb down including road runner's internal speed test. I hook up my
laptop to
>the same modem, same port, using the same cable and I get 17mb down. I have
>ruled out iptables, nics (because same result on both nics), cables, cable
>modem(with a brighthouse rep) any suggestions would be greatly appreciated.
>Thanks.
>
>**********************************************
>JohnStanley Writes:
>
>Maybe trying the vender specific drivers for the Nic Card in question on the
>CentOS Server may solve the problem. Or a possibilty is if your Bonding or
>teaming the nics there may be configuration issues in the way you have them
>set up. Try just one Nic card by itself with out Bonding. You can use
>"miitool" to set or force the transfer rate of the nic in
question; see "man
>miitool" Since you do say you have dual nics try them both separtly
like I
>stated (Without The Bonding).
>
>JohnStanley
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

Not a laptop it is a server with dual nics mtu set at 1500, 1 nic directly to
the cable modem the other to my lan.
---------- Original Message ----------------------------------
From: "Joseph L. Casale" <JCasale at activenetwerx.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Mon, 8 Sep 2008 23:46:40 -0600
>>I have a server with dual nic's running centos 5.2 1 nic connects to
the
>>cable modem, which has 15m d and 2m up, but speed tests barely come back
at
>>5mb down including road runner's internal speed test. I hook up my
laptop to
>>the same modem, same port, using the same cable and I get 17mb down. I
have
>>ruled out iptables, nics (because same result on both nics), cables,
cable
>>modem(with a brighthouse rep) any suggestions would be greatly
appreciated.
>>Thanks.
>
>So I gather the nics aren't teamed from your wording, maybe MTU is set
sub
>optimal for your particular ISP? What OS on your laptop? Wrong MTU can have
>a significant effect such as this.
>
>jlc
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: MHR <mhullrich at gmail.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 10:52:24 -0700
>On Tue, Sep 9, 2008 at 8:18 AM, David Petruzzella <sektor at dsli.com>
wrote:
>> Thanks I read about that, but I am having a hard time forcing duplex, I
don't remember the exact error it was giving me at the moment but from what
I read it looks like a driver issue.
>
>Would you please follow this email list's conventions and bottom post?
>
>Thank you.
>
>mhr
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
Sure sorry about that at work and what not its the opposite. 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: "Joseph L. Casale" <JCasale at activenetwerx.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 12:40:48 -0600
>>Not a laptop it is a server with dual nics mtu set at 1500, 1 nic
directly to the cable modem the other to my lan.
>
>MTU has nothing to do with laptops, it has to do with TCP/IP.
>Aside from a previous posters duplex setting (good catch on his part) its
>possible that MTU is an issue. Some ISP's do some silly stuff. You
>*are* going into a cable modem (which I read as accessing the internet
>through some means other than a high-end fiber/etc connection where oddities
>wouldn't likely exist:>)
>
>jlc
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
The MTU setting of 1500 is optimal for that connection from what I read and
according to tweak tests and I was just clarifying that the pc in question was
not a laptop I am aware that MTU is TCP related and I agree sometimes there are
funny things that happen with that, based on other tests I have done with other
devices kind of rules out MTU it seems to be more a duplexing issue or driver
issue I think.

Both myself and the cable guy have hooked a laptop to the cable modem all beit
they are windows laptops with the same settings and speed tests were normal.
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: Ric Moore <wayward4now at gmail.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 09 Sep 2008 14:56:58 -0400
>
>On Mon, 2008-09-08 at 23:46 -0600, Joseph L. Casale wrote:
>> >I have a server with dual nic's running centos 5.2 1 nic
connects to the
>> >cable modem, which has 15m d and 2m up, but speed tests barely come
back at
>> >5mb down including road runner's internal speed test. I hook up
my laptop to
>> >the same modem, same port, using the same cable and I get 17mb
down. I have
>> >ruled out iptables, nics (because same result on both nics),
cables, cable
>> >modem(with a brighthouse rep) any suggestions would be greatly
appreciated.
>> >Thanks.
>> 
>> So I gather the nics aren't teamed from your wording, maybe MTU is
set sub
>> optimal for your particular ISP? What OS on your laptop? Wrong MTU can
have
>> a significant effect such as this.
>
>"Back in the day..." there used to be a utility that would test
your
>connection for the optimal MTU setting. Is there something like that for
>Linux?? Ric
>
>-- 
>----------------------------------------------------
>My father, Victor Moore (Vic) used to say:
>"There are two Great Sins in the world...
>..the Sin of Ignorance, and the Sin of Stupidity.
>Only the former may be overcome." R.I.P. Dad.
>Linux user# 44256 Sign up at: http://counter.li.org/
>https://nuoar.dev.java.net/
>Verizon Cell # 336-254-1339
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
I got my information about the MTU setting from doing google searches and the
message boards on broadband reports as well as doing the tweak test on broad
band reports.
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: Scott Silva <ssilva at sgvwater.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 09 Sep 2008 12:22:53 -0700
>This is an OpenPGP/MIME signed message (RFC 2440 and 3156)
>
I agree but this is a newer ambit modem and if it was the case then mine and the
cable guy's laptop would experience the same results.




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: "RobertH" <roberth at abbacomm.net>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 12:32:33 -0700
>> 
>> I got my information about the MTU setting from doing google searches
and
>> the message boards on broadband reports as well as doing the tweak test
on
>> broad band reports.
>> 
>> 
>
>It appears you are on Comtrash where everything is CoMtRaSHtIc.
>
>Good for you. Dance a jig.
>
>Stop beating the MTU horse to death and get busy finding out how to best
>connect to the CPE devices.
>
>Google that stuff.
>
>First, when testing, disable (admin down) the internal network on the second
>nic
>
>Also, check to see what specific nics you have an if any recent issues with
>kernels and such and any updates.
>
>Google that too.
>
>Then Focus on the internet transit/transport infrastructure.
>
>Find out best way to interface with the cable modem CPE infrastructure.
>
>Meaning what speed and duplex.
>
>Trial and error.
>
>You will get there...
>
>Some nics need to be played with and dealt with in terms of kernel and other
>things like disable ipv6 in modprobe.conf and dhcp...
>
>You might be better off having an intermediate device between you and the
>cable modem if your nics don't wanna play nice with the cable modem.
>
>Last but not least, you can always check stats and see if collisions and
>errors on interfaces and more.
>
>Enjoy
>
> - rh
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
1. Comcrap is not available in my area I am on road runner biz through bright
house with 15m down and 2 up and static ip.

2.  Already tested without the internal nic running

3. Tried with a router in between and the result is the same.

4. I do agree about the negotiation aspect, which I'm looking into as well.

5. Don't notice any collisions or errors on ifconfig.
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: "RobertH" <roberth at abbacomm.net>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 13:19:41 -0700
>> 
>> 1. Comcrap is not available in my area I am on road runner biz through
>> bright house with 15m down and 2 up and static ip.
>> 
>> 2.  Already tested without the internal nic running
>> 
>> 3. Tried with a router in between and the result is the same.
>> 
>> 4. I do agree about the negotiation aspect, which I'm looking into
as
>> well.
>> 
>> 5. Don't notice any collisions or errors on ifconfig.
>> 
>> 
>
>I see, when I trace the domain you are sending email from I assumed it was
>you on a comcast link.
>
>Or maybe roadrunner resells comcast in your area or comcast bought
>roadrunner in your area?
>
>The dlsi.com dig pointing at 0.0.0.0 was interesting.
>
>Never ever seen dns point to 0.0.0.0
>
>Shouldn't there be a error catch in dns server software for that
situation?
>
>It's like using dns to route? Har har he.
>
>Or did you buy the internet?
>
>;-)
>
>Man, if so, I wanna be your best bud cause I need a piece of the action to
>sell at a good deal
>
>;->
>
> - rh
>
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
LOL I wish I bought the internet, no this dsli.com e-mail address is because I
used to have dsl service with them when they were a national provider back when
there were clec's, but when all the clec's went poof they could not keep
the dsl side of it, but as a sign of gratitude they let me and my wife keep the
e-mail account.  My domain is actually pbfamily.net
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: Les Mikesell <lesmikesell at gmail.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 09 Sep 2008 15:46:53 -0500
>David Petruzzella wrote:
>> 
>> I got my information about the MTU setting from doing google searches
and the message boards on broadband reports as well as doing the tweak test on
broad band reports.
>
>MTU's need to match the lowest value in the path between endpoints not 
>just your adjacent connection.  If there are tunnels/PPOE that add 
>encapsulating headers involved you'll have a performance hit when the 
>routers need to fragement and reassemble large packets.
>
>-- 
>   Les Mikesell
>    lesmikesell at gmail.com
>
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
No tunnels.
 




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: Scott Silva <ssilva at sgvwater.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 09 Sep 2008 14:00:01 -0700
>This is an OpenPGP/MIME signed message (RFC 2440 and 3156)
>
I tried that and I don't think I got ethtool to respond properly, I will
have to try again later.




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: "RobertH" <roberth at abbacomm.net>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 9 Sep 2008 12:36:11 -0700
>> 
>> I got my information about the MTU setting from doing google searches
and
>> the message boards on broadband reports as well as doing the tweak test
on
>> broad band reports.
>> 
>
>David
>
>I also noticed that if you
>
>dig dlsi.com
>
>it appears to come up as 0.0.0.0
>
>huh?
>
>Good for you.
>
>maybe you should change that depending on how you have things setup
>
>could be that your packets get a lil confused eh?
>
> - rh
>
>_______________________________________________
>CentOS mailing list
>CentOS at centos.org
>http://lists.centos.org/mailman/listinfo/centos
>
dsli.com is just for e-mail has nothing to do with my current internet service,
dsli used to be a national isp I had service with before all the clecs went poof
and as a sign of gratitude they let me and my wife keep our e-mail accounts.




________________________________________________________________
Sent via the WebMail system at mail.dsli.com

---------- Original Message ----------------------------------
From: Scott Silva <ssilva at sgvwater.com>
Reply-To: CentOS mailing list <centos at centos.org>
Date:  Tue, 09 Sep 2008 14:07:24 -0700
>This is an OpenPGP/MIME signed message (RFC 2440 and 3156)
>
who knows its been a long time since I talked with anyone at my old isp
haven't had them for years.




________________________________________________________________
Sent via the WebMail system at mail.dsli.com