On Mon, April 10, 2017 4:17 pm, John R Pierce wrote:> On 4/10/2017 1:57 PM, m.roth at 5-cent.us wrote: >> In what universe are those "consistant" device names, as opposed to >> eth[0...]? And how could it help automated scripts that you can run on >> *any* system you're administering? > > if I have a Intel gigE interface and a Marvell 10g interfaces, which one > is eth0 and why? > > Say its Intel on eth0 and Marvell on eth1, if I then add another intel, > is the Marvell now eth2 ? >Without intent to contradict... I really would prefer them numbered according to their bus address. Not in the order (or reverse order - as it was once) of them been discovered. And if you add hardware with bus address between those of eth0 and eth1, you will have newly added one become eth1, and former eth1 becomes eth2. I know, it stems from old idiotic habit to always look inside the boxes... call me an old UNIX outcast. (No, don't, that would be a complement I unlikely deserve ;-) Valeri> > -- > john r pierce, recycling bits in santa cruz > > _______________________________________________ > CentOS mailing list > CentOS at centos.org > https://lists.centos.org/mailman/listinfo/centos >++++++++++++++++++++++++++++++++++++++++ Valeri Galtsev Sr System Administrator Department of Astronomy and Astrophysics Kavli Institute for Cosmological Physics University of Chicago Phone: 773-702-4247 ++++++++++++++++++++++++++++++++++++++++
On 4/10/2017 2:27 PM, Valeri Galtsev wrote:> Without intent to contradict... I really would prefer them numbered > according to their bus address. Not in the order (or reverse order - as it > was once) of them been discovered. And if you add hardware with bus > address between those of eth0 and eth1, you will have newly added one > become eth1, and former eth1 becomes eth2. I know, it stems from old > idiotic habit to always look inside the boxes... call me an old UNIX > outcast. (No, don't, that would be a complement I unlikely deservebus address is a somewhat nebulous concept on PCI, afaik, its not neccessarily slot specific. its even messier on things like USB -- john r pierce, recycling bits in santa cruz
And to add to the issues concerning consistency, what if you have a fail over unit and you're replicating configuration (i.e. NIC), you really prefer that NIC names remain the same. Given the hardware morass underneath all of this, the only safe choice (regardless of whether you use systemd, init scripts, whatever) is to take control and force the association of the name to the MAC address in the appropriate file. It doesn't allow total configuration replication but it does allow enough (with judicious use of segregation) to be useful. This does concern me, another post referred to the heavy-handed way in which systemd has been implemented and I totally agree. "You will conform" - no exceptions. What I fear is that we will lose the ability to control the name, MAC address association at some future point because "no one needs to do that" (speaking from their ivory tower). ----- Original Message ----- From: "John R Pierce" <pierce at hogranch.com> To: "centos" <centos at centos.org> Sent: Monday, April 10, 2017 4:36:48 PM Subject: Re: [CentOS] OT: systemd Poll On 4/10/2017 2:27 PM, Valeri Galtsev wrote:> Without intent to contradict... I really would prefer them numbered > according to their bus address. Not in the order (or reverse order - as it > was once) of them been discovered. And if you add hardware with bus > address between those of eth0 and eth1, you will have newly added one > become eth1, and former eth1 becomes eth2. I know, it stems from old > idiotic habit to always look inside the boxes... call me an old UNIX > outcast. (No, don't, that would be a complement I unlikely deservebus address is a somewhat nebulous concept on PCI, afaik, its not neccessarily slot specific. its even messier on things like USB -- john r pierce, recycling bits in santa cruz _______________________________________________ CentOS mailing list CentOS at centos.org https://lists.centos.org/mailman/listinfo/centos