Hi all, I swapped a Realtek 8139 100M Net Card for a Realtek 8169 Gig Net card in a Centos 5.6 system.
After reconfiguring the network settings I find that on boot up the 8169 interface does not start. The startup script says something like "The 8169 is not available..."
ifconfig -a shows a device with the right mac address called __tmpsomethingorother
I can get it going if I remod the r8169 driver and load it again. Then ifup eth1 works.
There's also a Marvell 88E8056 on the motherboard. /var/log/messages shows the 8169 as eth0 and the Marvell as eth1 but when the network is finally up they are the other way round. ifcfg-eth0 has the mac address of the Marvell in it and ifcfg-eth1 doesn't have a mac address in it.
Should I wipe those scripts and start again?
I've put a nasty bodge in rc.local for the moment to make the startup work.
Thanks
Ken
--On Friday, March 16, 2012 07:41:18 PM +0000 Ken Smith kens@kensnet.org wrote:
Hi all, I swapped a Realtek 8139 100M Net Card for a Realtek 8169 Gig Net card in a Centos 5.6 system.
Although the 100M versions were passable, I found the GB versions to be so flakey to be not worth it; I ripped them out and replaced them with Intel cards, which have been solid.
In my case, I found that under load they would just "go away" ... become unresponsive and need a power cycle to clear them. (It didn't appear to be an autoneg issue.)
In some cases, they didn't get recognised at boot, either, whereas a subsequent reboot *might* make them come back.
Devin
On 03/16/2012 04:03 PM, Devin Reade wrote:
--On Friday, March 16, 2012 07:41:18 PM +0000 Ken Smith kens@kensnet.org wrote:
Hi all, I swapped a Realtek 8139 100M Net Card for a Realtek 8169 Gig Net card in a Centos 5.6 system.
Although the 100M versions were passable, I found the GB versions to be so flakey to be not worth it; I ripped them out and replaced them with Intel cards, which have been solid.
In my case, I found that under load they would just "go away" ... become unresponsive and need a power cycle to clear them. (It didn't appear to be an autoneg issue.)
In some cases, they didn't get recognised at boot, either, whereas a subsequent reboot *might* make them come back.
Devin
I've seen the same behavious. Realtek cards are no longer an option for my builds. I would second Devin and suggest saving yourself a lot of headaches... Replace with an Intel NIC. The difference in performance and reliability is staggering.
Digimer wrote:
On 03/16/2012 04:03 PM, Devin Reade wrote:
--On Friday, March 16, 2012 07:41:18 PM +0000 Ken Smith kens@kensnet.org wrote:
Hi all, I swapped a Realtek 8139 100M Net Card for a Realtek 8169 Gig Net card in a Centos 5.6 system.
<snip>
In some cases, they didn't get recognised at boot, either, whereas a subsequent reboot *might* make them come back.
I've seen the same behavious. Realtek cards are no longer an option for my builds. I would second Devin and suggest saving yourself a lot of headaches... Replace with an Intel NIC. The difference in performance and reliability is staggering.
If the OP doesn't/can't go out to buy another NIC, you might try putting HWADDR in ifconfig-eth<whatever>. If it was CentOS 6.x, you'd have to deal with /etc/udev/rules.d/70-persistent-net-rules
mark "just had this happen"
On Fri, 2012-03-16 at 19:41 +0000, Ken Smith wrote:
After reconfiguring the network settings I find that on boot up the 8169 interface does not start. The startup script says something like "The 8169 is not available..."
ifconfig -a shows a device with the right mac address called __tmpsomethingorother
The __tmpsomethingoroth names are used by udev to shuffle the device names (like when you set HWADDR in the conf file). Looks like it saw enough of the device to start the process but the device "disappeared" before it could finish. Had this happen with an onboard Gbit Realteak NIC that was slowly dying. Rebooting would sometimes get it to work. After a week of mucking around with setting HWADRR and writing udev rules, it eventually died completely. A new card would have been a lot cheaper than my time.
Kal