On Tue, Jul 29, 2014 at 5:57 AM, SCHAER Frederic <frederic.sch...@cea.fr> wrote: >> De : De la part de Tom H >> Envoyé : lundi 28 juillet 2014 23:35 >> On Mon, Jul 28, 2014 at 5:35 AM, SCHAER Frederic <frederic.sch...@cea.fr> >> wrote: >>> >>> From time to time, we get reboot issues with some machines, and each time it >>> looks like there are duplicated persistent rules for the Ethernet devices : >>> >>> cat /etc/udev/rules.d/70-persistent-net.rules >>> >>> # PCI device 0x8086:0x1521 (igb) (custom name provided by external tool) >>> SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", >>> ATTR{address}=="84:8f:69:fb:c1:2a", ATTR{type}=="1", KERNEL=="eth*", >>> NAME="em1" >>> >>> # PCI device 0x8086:0x1521 (igb) (custom name provided by external tool) >>> SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", >>> ATTR{address}=="84:8f:69:fb:c1:2b", ATTR{type}=="1", KERNEL=="eth*", >>> NAME="em2" >>> >>> # PCI device 0x8086:0x1521 (igb) (custom name provided by external tool) >>> SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", >>> ATTR{address}=="84:8f:69:fb:c1:2a", ATTR{type}=="1", KERNEL=="eth*", >>> NAME="eth0" >>> >>> # PCI device 0x8086:0x1521 (igb) (custom name provided by external tool) >>> SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", >>> ATTR{address}=="84:8f:69:fb:c1:2b", ATTR{type}=="1", KERNEL=="eth*", >>> NAME="eth1" >>> >>> The second set of rules seem to overwrite the first one, and then we get >>> issues with our network config. >>> >>> This does not happen on all nodes, just apparently to some random ones, >>> sometimes. >>> >>> I’m wondering if some of you might have faced and solved that erratic thing >>> already ? >>> >>> We want to keep the emX scheme for nodes which support it… >> >> The emX names come from biosdevname (part of a base install): >> >> https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/6/html/Deployment_Guide/appe- >> Consistent_Network_Device_Naming.html >> >> You should be able to delete the last two rules, the ethX rules, for >> your interfaces to use the emX names, unless you're using >> "biosdevname=0" on the kernel cmdline. > > Thanks for your reply. > > That's actually what I do to recover a happy system, but I'd like to prevent > this from hapenning. > I don't want to force ethX when nodes support emX, but when I use emX, I > don't want neither the system to suddenly tell me emX became ethX, which is > what happens. > > What I'm wondering is why do I suddenly have both the emX and ethX devices > in udev, despite the fact the node has lived for days with emX only, rebooted > correctly without changing anything...? > > Firmware bug ? (I know emX scheme depends on firmwares and DMI information...)
The only time that I've seen "write_net_rules" write a new rule in 70-persistent-net.rules is when there's a error in the existing rule, usually an incorrect mac address. But this isn't the case here so either you've hit a bug or biosdevname was disabled on your system either by using "biosdevname=0" on the kernel cmdline or by uninstalling biosdevname. To ensure that this doesn't happen: mv /etc/udev/rules.d/70-persistent-net.rules /etc/udev/rules.d/90-netX-persistent-net.rules ln -s /dev/null /etc/udev/rules.d/70-persistent-net.rules