Anyone got any suggestions for this.
I have a linksys sr2024 Gb switch
Several things are plugged into this - all using cat 5e or cat 6 cables.
Some things are connecting with full Gb speed but the main firewall wont.
The machine has 3 intel nic's -
lspci | egrep -i ethernet 04:00.0 Ethernet controller: Intel Corporation 82573L Gigabit Ethernet Controller (rev 01) 05:04.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05) 05:05.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)
The 82573 is on board and not used - the other 2 nic's are used - one for internal subnet and one for external.
The internet side is connected to a cisco 800 router (which I believe is 10/100 only .. tho I cudnt swear for sure)
The inside is connected to the SR2024 Gb switch.
Both nic's come up in 100 Mb/s - I rebooted the switch, and I plugged and unplugged the cables, I tried ethtool -r ethx a few times to no avail.
Other machines on same switch are happily running at full gigabit speeds.
Nothing in logs suggests a problem
Is there a problem with multi nics of same brand where they need to run at diff speeds ? Or is this a quirk with some hardware somewhere ? Or ???
Appreciate any suggesions ... its getting tiring forcing renogiation!
thanks ...
gene/
On 09/18/2010 06:08 PM, Genes MailLists wrote:
Anyone got any suggestions for this.
I have a linksys sr2024 Gb switch
Several things are plugged into this - all using cat 5e or cat 6 cables.
Some things are connecting with full Gb speed but the main firewall wont.
The machine has 3 intel nic's -
lspci | egrep -i ethernet 04:00.0 Ethernet controller: Intel Corporation 82573L Gigabit Ethernet Controller (rev 01) 05:04.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05) 05:05.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)
The 82573 is on board and not used - the other 2 nic's are used - one for internal subnet and one for external.
The internet side is connected to a cisco 800 router (which I believe is 10/100 only .. tho I cudnt swear for sure)
The inside is connected to the SR2024 Gb switch.
Both nic's come up in 100 Mb/s - I rebooted the switch, and I plugged and unplugged the cables, I tried ethtool -r ethx a few times to no avail.
Did you try ethtool -s ethX speed 1000
Other machines on same switch are happily running at full gigabit speeds.
Nothing in logs suggests a problem
Is there a problem with multi nics of same brand where they need to run at diff speeds ? Or is this a quirk with some hardware somewhere ? Or ???
Appreciate any suggesions ... its getting tiring forcing renogiation!
thanks ...
gene/
On Sat, 2010-09-18 at 21:08 -0400, Genes MailLists wrote:
Anyone got any suggestions for this.
I have a linksys sr2024 Gb switch
Several things are plugged into this - all using cat 5e or cat 6 cables.
Some things are connecting with full Gb speed but the main firewall wont.
The machine has 3 intel nic's -
lspci | egrep -i ethernet 04:00.0 Ethernet controller: Intel Corporation 82573L Gigabit Ethernet Controller (rev 01) 05:04.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05) 05:05.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)
The 82573 is on board and not used - the other 2 nic's are used - one for internal subnet and one for external.
The internet side is connected to a cisco 800 router (which I believe is 10/100 only .. tho I cudnt swear for sure)
The inside is connected to the SR2024 Gb switch.
Both nic's come up in 100 Mb/s - I rebooted the switch, and I plugged and unplugged the cables, I tried ethtool -r ethx a few times to no avail.
Other machines on same switch are happily running at full gigabit speeds.
Nothing in logs suggests a problem
Is there a problem with multi nics of same brand where they need to run at diff speeds ? Or is this a quirk with some hardware somewhere ? Or ???
Unlikely. We use Intel NIC's (both 1GbE and 10GbE) extensively (up to 16 linkes per machine) for link monitoring without issues.
Appreciate any suggesions ... its getting tiring forcing renogiation!
Try connecting the two links to each other. Set their respective IP's to 1.1.1.1/255.255.255.0 (eth1) and 2.2.2.2/255.255.255.0 (eth2). Now, check that they both auto negotiate to 1GbE/Full duplex. If they don't, you have a wire issue. (Is it CAT 5 or above?) If they do, do the following:
1. Install iptraf. ($ yum install iptraf -y) 2. Launch iptrag and configure it to work in promiscuous mode (Configure -> Force promiscuous mode). 3. Go back, select the "general interface statistics". 4. From another console, start flood pinging. ping -s 1400 -f -i eth2 -b 1.1.1.255 5. Return to the iptraf console, you should the see the packet increase linearly on both interfaces. If it doesn't, something is broken, if it does - you have a switch problem.
- Gilboa
On 19/09/10 15:18, Gilboa Davara wrote:
If they don't, you have a wire issue. (Is it CAT 5 or above?)
Does all your cables have 2 pairs of wires or 4 pairs of wires, and are they properly pinned out ?
Even in purchased cables, I have seen such faults (especially cheap/no name) cables, where the pins for 100Mb are correctly pinned out, but the other 4x wires are not.
Are they hand terminated, did the pins crimp down properly etc ? It's not uncommon to have faulty raw plugs, or crimp tools (eg that scar the edges of the plastic channels - making the copper fingers be unable to make contact).
A couple of users have reported problems with that specific model NIC which were caused by bad capacitors in the switch. You might look at replacing that or opening it up to see if you can spot any bad caps: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/309211
On 09/19/2010 01:18 AM, Gilboa Davara wrote:
Unlikely. We use Intel NIC's (both 1GbE and 10GbE) extensively (up to 16 linkes per machine) for link monitoring without issues.
Great to know thanks ..
Try connecting the two links to each other. Set their respective IP's to 1.1.1.1/255.255.255.0 (eth1) and 2.2.2.2/255.255.255.0 (eth2). Now, check that they both auto negotiate to 1GbE/Full duplex. If they don't, you have a wire issue. (Is it CAT 5 or above?) If they do, do the following:
- Install iptraf. ($ yum install iptraf -y)
- Launch iptrag and configure it to work in promiscuous mode (Configure
-> Force promiscuous mode). 3. Go back, select the "general interface statistics". 4. From another console, start flood pinging. ping -s 1400 -f -i eth2 -b 1.1.1.255 5. Return to the iptraf console, you should the see the packet increase linearly on both interfaces. If it doesn't, something is broken, if it does - you have a switch problem.
Great idea ... I will try this when I'm able ..
On 09/19/2010 01:34 AM, David Timms wrote:
Does all your cables have 2 pairs of wires or 4 pairs of wires, and are they properly pinned out ?
The cat 5e's were punched by my networking electrician and checked with a (fairly expensive) network tester. The cat 6's were purchased (550 Mhz) premade cables ... they all have 4 pairs ..
I've havent seen 2 pairs of wires .. mmm at least since pre cat 3 days ... :-)
Agreed dodgy cables can be an issue ... for good measure I had flipped a few cables ... same result.
Thanks
On 09/19/2010 03:34 AM, Gordon Messmer wrote:
A couple of users have reported problems with that specific model NIC which were caused by bad capacitors in the switch. You might look at replacing that or opening it up to see if you can spot any bad caps: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/309211
Interesting thread indeed ... very similar to what I'm seeing .. soem of the bad caps seemed to be not associated with a specific port, suggesting power supply levels on board perhaps ...
Thanks for the tip!
g
On 09/19/2010 01:18 AM, Gilboa Davara wrote:
On Sat, 2010-09-18 at 21:08 -0400, Genes MailLists wrote:
The machine has 3 intel nic's -
lspci | egrep -i ethernet 04:00.0 Ethernet controller: Intel Corporation 82573L Gigabit Ethernet Controller (rev 01) 05:04.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05) 05:05.0 Ethernet controller: Intel Corporation 82541PI Gigabit Ethernet Controller (rev 05)
..
Both nic's come up in 100 Mb/s - I rebooted the switch, and I plugged and unplugged the cables, I tried ethtool -r ethx a few times to no avail.
Other machines on same switch are happily running at full gigabit speeds.
Nothing in logs suggests a problem
Is there a problem with multi nics of same brand where they need to run at diff speeds ? Or is this a quirk with some hardware somewhere ? Or ???
Unlikely. We use Intel NIC's (both 1GbE and 10GbE) extensively (up to 16 linkes per machine) for link monitoring without issues.
I have confirmed that when the 2 intel 8254 NIC's are in play - they both run at 100 - when i switch out one of the and instead use one 8254 and the other 8257 (also intel) then indeed the one connected to the Gb switch runs at 1 Gb and the other remains at 100.
So this conforms a bug - there is a problem with 2 identical NICs both running at 100 instead of 1 at a Gb (for hardware described above anyway)
Now this is an older kernel (2.6.27) so it may have been fixed since then - anyone know ?
gene/
-
Development discussions related to Fedora -
Genes MailLists -
Gilboa Davara -
Gordon Messmer -
JD