Differences between revisions 2 and 66 (spanning 64 versions)
Revision 2 as of 2019-09-11 18:39:54
Size: 12995
Editor: JustinBRye
Comment: fixing up duff markup
Revision 66 as of 2019-09-18 09:13:49
Size: 21332
Editor: JustinBRye
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
This is only a first draft, but the plan is to turn it into something
that the Release Notes can point at as the nearest thing we've got to
an official guide.

todo: links from

 * NetworkConfiguration (replacing the halfbaked version there)

 * NetworkFAQ (much of that page needs replacing)

<<TableOfContents()>>

'''Warning''': Anything that changes the names of your network
#language en

'''See below''' for a [[#migration|"Predictable Names" Migration HOWTO]]. /* (Comments on) */

{{{#!wiki warning
Anything that changes the names of your network
Line 17: Line 10:
}}}

This page deals with the various schemes by which wired and wireless network interfaces are assigned names - that is, the underlying system labels like `eth0` or `wlx800e1319c734`. It has nothing to do with the "connection profile" names used by apps such as NetworkManager, like "Wired connection 1".

<<TableOfContents()>>

----
Line 26: Line 26:
At least in theory, if module probes completed in a different order,
`eth0` and `eth1` might switch places on successive boots. As boot
processes became less linear and interfaces became more hotpluggable
this became more of a concern.
At least in theory, if module probes completed in a different order, `eth0` and `eth1` might switch places on successive boots. As boot processes became less linear and interfaces became more hotpluggable this became more of a concern.
Line 33: Line 30:
If you wipe out all other mechanisms (booting with the `ifnames`
mechanism deactivated and no
`/etc/udev/rules.d/75-persistent-net.rules` file, or maybe just a
`udev` that has finally stopped supporting it) then I suppose you'll
be back to this one.

(Is this account missing old approaches to interface-renaming that
still need to be dealt with? For instance, does the package
[[DebianPkg:ifrename|ifrename]] still work?)
If you wipe out all other name-assignment mechanisms then you'll be left with this one.

The simple way of disabling the whole [[#predictable|current interface naming scheme]] (which you might want to try for one-off testing) is just to boot with the kernel parameter `net.ifnames=0`, which can be set in an interactive `grub` session at boot or made persistent by editing `/etc/default/grub` and running `update-grub`.

Alternatively, you can override `/lib/systemd/network/99-default.link`, with a custom version in `/etc/systemd/network/`, '''or''' similarly override `/lib/udev/rules.d/80-net-setup-link.rules`, '''or''' mask the latter by using a `/dev/null` symlink instead of a custom version, or... there seem to be lots of ways of doing this, so make sure you haven't done it in more than one way or it'll trip you up in a couple of years when you try to undo it. See the [[#ref|external links]] below on standard methods for overriding systemd configuration. Oh, and beware of [[#initrd|initrd skew]].

(If you've still got a working [[#legacy|legacy]] `70-persistent-net.rules` file, the `net.ifnames=0` flag doesn't deactivate that, so you'd need to see [[#trigger|the instructions below]] for temporarily disabling that file. It's probably also possible to do this by masking enough of systemd.)

----
Line 44: Line 41:
= THE "PERSISTENT NAMES" APPROACH =

This scheme, introduced somewhere around Debian 5 "Lenny", used `udev`
to identify interfaces by MAC address and assign a fixed interface
number to any interface it recognized (writing the rules to
`/etc/udev/rules.d/75-persistent-net.rules`). This could have annoying
side-effects (e.g. if you were replacing a machine's sole NIC, you'd
also have to take special care to ensure it took over as network
interface number 0) but these were minor and predictable.
= THE "PERSISTENT NAMES" SCHEME =

This scheme, introduced somewhere around Debian 5 "Lenny", used `udev` to identify interfaces by MAC address and assign a fixed interface number to any interface it recognized (writing the rules to `/etc/udev/rules.d/70-persistent-net.rules`). This could have annoying side-effects (e.g. if you were replacing a machine's sole NIC, you'd also have to take special care to ensure it took over as network interface number 0) but these were minor and predictable.

A sample `.rules` entry:
{{{
 #/etc/udev/rules.d/70-persistent-net.rules
 # PCI device 0x8086:0x100e (e1000e)
 SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="01:23:45:67:89:ab", ATTR{type}=="1", KERNEL=="eth*", NAME="eth0"
}}}

Since Debian 9 "Stretch", newly installed machines no longer start with an `/etc/udev/rules.d/70-persistent-net.rules` file, though such files are maintained if they still exist (with new lines added for newly installed network hardware). Debian 10 "Buster" additionally lacks the `/lib/udev/rules.d/75-persistent-net-generator.rules` file that appends to it, though legacy `70-persistent-net.rules` files are still honored. This support is expected to be removed from Debian 11 "Bullseye" at some time before it is released. /* No further details have been given of just when or how we should expect this support to end; is `udev` going to stop honoring `.rules` files that set NAME= on network interfaces, or what? */
Line 56: Line 56:
This still had subtle race conditions, required `/etc` to be on a
writable file system, and had problems with virtualization, so it's no
longer supported upstream; the plan (still taken for granted in most
of the documentation) was for it not to be supported in Debian 10
"Buster", but in the end it has scraped through into another release
cycle.
This still had subtle race conditions, required `/etc` to be on a writable file system, and had problems with virtualization, so it's no longer supported upstream; the plan (still taken for granted in most of the documentation) was for it not to be supported in Debian 10 "Buster", but in the end it has scraped through into another release cycle.

<<Anchor(persistence)>>
Line 65: Line 61:
First, make sure you've got a working "legacy"
`/etc/udev/rules.d/75-persistent-net.rules` file. Then deactivate the
replacement scheme. The simple way of doing that latter (which you
might want to try for one-off testing) is just to boot with the kernel
parameter `net.ifnames=0`, which can be set in an interactive `grub`
session at boot (or made persistent by editing `/etc/default/grub` and
running `update-grub`).

Alternatively, you can override `/lib/systemd/network/99-default.link`,
with a custom version in `/etc/systemd/network/`, '''or''' similarly
override `/lib/udev/rules.d/80-net-setup-link.rules`, '''or''' mask
the latter by using a `/dev/null` symlink instead of a custom version,
or... there seem to be lots of ways of doing this, so make sure you
haven't done it in more than one way or it'll trip you up in a couple
of years when you try to undo it. Oh, and beware of [[#initrd|initrd
skew]].

Unfortunately, none of this will carry on working beyond the point
when `udev` stops supporting the legacy `75-persistent-net.rules`
mechanism, so you should be ready to switch to a different system when
that happens.

== How to disable it ==

If you've currently got a legacy
`/etc/udev/rules.d/75-persistent-net.rules` file but have decided to
switch to the [[#ifnames|new regime]], you can do that just by
disabling the `.rules` file; see the `udev` README.Debian.gz

<<Anchor(biosdevname)>>
= ANOTHER APPROACH SOMETIMES SEEN ON NON-DEBIAN SYSTEMS =

Old releases of RedHat (among others) used a "biosdevname" system,
but that's never been supported under Debian. If you need to know
about it there's bound to be documentation somewhere.

<<Anchor(ifnames)>>
= THE NEW REGIME =

The new "net.ifnames" approach uses names usually derived from the
location of the interface in terms of hardware buses etc.

(This system is often advertised as providing "Predictable Names",
but the main thing that's predictable about it is that calling it
this will cause furious users to pop up disputing the appropriateness
of that name. Can we just skip all that here, please?)
If you've got a working "legacy" `/etc/udev/rules.d/70-persistent-net.rules` file and want to stick with it, you can safely upgrade through Debian 9 "Stretch" and Debian 10 "Buster"; `udev` on these releases still respects that file if present (and will accept a freshly created one). However, bear in mind that on Buster you'll need to maintain its contents yourself, and eventually even this legacy support will end, so you should be ready to switch to a different scheme before that happens.

== How to let go and move on ==

If you're currently running something [[#jessie|newer than Debian 8 "Jessie"]] with a legacy `70-persistent-net.rules` file but have decided to switch to the [[#predictable|new regime]], you can do that just by disabling the `.rules` file (then [[#initrd|updating the initrd]] before you reboot); see the `udev` README.Debian.gz and [[#migration|the more detailed guide below]].

----

<<Anchor(misc)>>
= MISCELLANEOUS OLD SCHEMES =

(Contributions welcome)

Several workarounds for renaming interfaces grew up in the early days of hotpluggable wireless interfaces, but if they still work it'll be because like [[DebianPkg:ifrename]] they now use `udev` rules under the hood. It's not clear what remaining advantage this has over [[#custom|the canonical .link approach]] - is it perhaps useful for non-systemd machines?

Old releases of !RedHat (among others) used a "biosdevname" system, but that's never been supported under Debian. If you need to know about it there's bound to be documentation somewhere.

----

<<Anchor(predictable)>>
= THE "PREDICTABLE NAMES" SCHEME =

The new scheme uses names usually derived from the location of the interface in terms of hardware buses etc: `eno1`, `wlp1s3`. The idea was that this provides "Predictable Names", though as it turns out the main thing that's predictable about it is that calling it this will cause furious users to pop up disputing the appropriateness of that name. (Can we just skip all that here, please?) /* Obviously they should have called them "deterministically computable network interface names"... */
Line 114: Line 87:
This should be easy enough; before you start configuring firewalls
etc., just look at the output of (e.g.)

{{{
        ip a
}}}

Unlike the old days, when the only way to guess which cable was
plugged into `eth0` and which was `eth1` was to keep track of MAC
addresses, this system provides extra clues in the interface names.
/* Either `enp0s1` is the one on the left and `enp1s0` is the one on
the right, or equally likely it's the other way round. See? So much
easier. */

== How to switch ==

It's advisable to do this as a separate thing in its own right, not
as part of a general dist-upgrade. However, if your PC only has one
network interface and not much is at stake you can try:

=== STRATEGY A ===

 * wait until `udev` breaks your networking, if it's going to, or
trigger the change by disabling the persistent-names system.

 * ask `ip` what new name it's using, and fix your config files.

You should probably at least check in advance to see what files
hard-code interface names, by running something like

{{{
        sudo rgrep ''wlan0'' /etc
}}}

(Obvious possibilities include `/etc/network/interfaces` and
configuration files for firewalls, wifi, DHCP...)

=== STRATEGY B ===

This strategy, more or less compulsory for remote servers, runs along
the lines of:
This should be easy enough; before you start configuring firewalls etc., just look at (e.g.) the output of `ip a` and note the names of the interfaces. Unlike the old days, when the only way to guess which cable was plugged into `eth0` and which was `eth1` was to keep track of MAC addresses, this system provides extra clues in the interface names. /* Either `enp0s1` is the one on the left and `enp1s0` is the one on the right, or equally likely it's the other way round. See? So much easier. */

<<Anchor(migration)>>
== How to migrate to this scheme on upgraded systems ==

It's advisable to do this as a separate migration in its own right, not as part of a general distribution upgrade. However, if your PC only has one network interface and not much is at stake you can try:

'''Strategy A'''

 * wait until `udev` breaks your networking, if it's going to, or [[#trigger|trigger]] the change yourself.

 * ask `ip` what new name it's using, and fix your configuration files.

You should probably at least check in advance to see what files hard-code interface names, by running something like

{{{
        sudo rgrep wlan0 /etc
}}}

Obvious likely hits include `/etc/network/interfaces` and configuration files for firewalls, wifi, DHCP... but it's possible that (e.g.) a laptop with a single wifi interface managed by !NetworkManager might need no fixing at all. /* That's only a guess, though - input from NM users welcome */

'''Strategy B'''

This strategy, more or less compulsory for remote servers, runs along the lines of:
Line 158: Line 114:
 * consider a dead-man's-handle `etckeeper` cronjob or the like

 * consult the list of [[#corner|corner cases]] below and search the
Internet for others
 * consider a dead-man's-handle `etckeeper` cronjob [[#nets|or the like]]

 * consult the list of [[#corner|corner cases]] below and search the Internet for others
Line 167: Line 122:
To find out what names `udev` would be choosing between if you switched
over to the new system, first get a list of the network devices the
system knows about:
To find out what names `udev` would be choosing between if you switched over to the new system, first get a list of the network devices the system knows about:
Line 175: Line 128:
For each device (other than `lo`), ask `udevadm` what NET_IDs it knows:

{{{
        udevadm test-builtin net_id /sys/class/net/''enp0s1'' 2>/dev/null
}}}

It's likely to tell you about things like ID_OUI_FROM_DATABASE and an
ID_NET_NAMING_SCHEME, but the lines that matter are the ones (given
in unhelpfully
random order) starting with ID_NET_NAME_. One of these
is the name that `udev` will give priority to - the list of candidates
may be so short that all you need to know is that ..._PATH beats
..._MAC, but there are also some rarer possibilities, and in general
if something unusual shows up then it will take priority.
For each device path (other than `/sys/class/net/lo`), ask `udevadm` what NET_IDs it knows:

{{{
        udevadm test-builtin net_id /sys/class/net/enp0s1 2>/dev/null
}}}

It's likely to tell you about things like ID_OUI_FROM_DATABASE and an ID_NET_NAMING_SCHEME, but the lines that matter are the ones (given in random order) starting with ID_NET_NAME_. One of these is the name that `udev` will give priority to - the list of candidates may be so short that all you need to know is that ..._PATH beats ..._MAC, but there are also some rarer possibilities, and in general if something unusual shows up then it will take priority. /* You'd think in a sane world we'd just ask for `udevadm ifnames` and it would list the interfaces it knows with their available names ordered by the priority policy currently in force... or, wait, surely in a ''sane'' world we'd be free to use any of these names, the way we can use any of the aliases in `/dev/disks` to refer to a hard disk. But that would be too simple. */
Line 191: Line 138:
 * ID_NET_NAME_FROM_DATABASE=:: very rare (possibly mythical) and
not to be confused with ID_OUI_FROM_DATABASE; if present, it gets
maximum priority.
Nobody seems to know why - it's mentioned in
[[DebianMan:5/systemd.link|systemd.link(5)]], but otherwise
undocumented. The database is hardcoded into `udev` and has only one
known entry, the spooky-sounding `idrac`.
/* My theory is that the global vampire conspiracy set this up so that
we've technically already invited them to cross the threshold. */

 * ID_NET_NAME_ONBOARD=:: appears for some but not all kinds of
onboard network card (Ethernet only, or wireless too?) - it's usually
a nice simple name like `eno0`.

 * ID_NET_NAME_SLOT=:: appears for some PCI-hotplug cards. Usually
looks like `ens0` (again, any wifi cases? Does it ever occur along
with
_ONBOARD?)

 * ID_NET_NAME_PATH=:: always present; usually something just
complicated enough to be easy to forget, like `wlp3s5` or `enp1s3f0`.
Note that all numbers are in hex.

 * ID_NET_NAME_MAC=:: also always present, but with a low enough
priority that by default it won't be used; e.g. `wlx800e1319c734`

You'd think in a sane world we'd be able to just ask for

{{{
        udevadm ifnames
}}}

and it would list the interfaces it knows with their available names
ordered by the priority policy currently in force... or, wait, surely
in a ''sane'' world we'd be free to use any of these names, the way we
can use any of the aliases in `/dev/disks` to refer to a hard disk.
But no, that would be too simple.
<<Anchor(hierarchy)>>
 1. '''
ID_NET_NAME_FROM_DATABASE=''' Very rare and not to be confused with ID_OUI_FROM_DATABASE; if present, it outranks any other ID_NET_NAME. The database is hardcoded into `udev` and has only one known entry, the spooky-sounding `idrac`. /* My theory is that the global vampire conspiracy set this up so that we've technically already invited them to cross the threshold. */

 1. '''ID_NET_NAME_ONBOARD=''' Appears for some but not all kinds of onboard network card - it's usually a nice simple name like `eno0` or `wlo0`.

 1. '''ID_NET_NAME_SLOT=''' Appears for some PCI-hotplug cards. Usually looks like `ens0` or `wls0`. (Does this ever occur alongside _ONBOARD?)

 1. '''ID_NET_NAME_PATH=''' Always present; usually something just complicated enough to be easy to forget, like `wlp3s5` or `enp1s3f0`. Note that all numbers are in hex.

 1. '''ID_NET_NAME_MAC=''' Also always present, but with a low enough priority that by default it won't be used; e.g. `wlx800e1319c734`

One good reason for doing the migration separately from a dist-upgrade is that the answers you get from (e.g.) Stretch's `udevadm` aren't absolutely guaranteed to be identical to the answers you get from Buster's. This further implies that any time you update `systemd` and reboot there's a chance your server might fall off the net, which seems like a good argument for using a [[#custom|customized scheme]] at least for the interface you're SSHing in on.
Line 228: Line 152:
== Corner cases ==

(Please try to avoid ballooning this section with tales of "I
don't know how this happened but it all went wrong for me"...)

 * TRIGGERING THE SWITCH: you don't need to ''delete''
`/etc/udev/rules.d/70-persistent-net.rules` to test what network
interface names you'll get without it. Just renaming it (e.g. to
`70-persistent-net.rules.old`) or commenting out particular lines
should be enough (as long as you update your [[#initrd|initramfs]]
before you reboot).

 * THE ID_NET_NAME_ HIERARCHY:: if you're ignoring
ID_NET_NAME_BORING on the assumption that anything you don't
understand probably isn't important, you need to reread the above -
the general rule is, if you don't recognise it, it'll mess you up.
== Complications and corner cases ==

(Additions welcome, but please try to avoid ballooning this section with tales of "I don't know how this happened but it all went wrong for me"...)

<<Anchor(jessie)>>
 * ANTIQUE SYSTEMS:: on Debian 9 "Stretch" or newer, merely booting without a `net.ifnames=0` override (and without a `70-persistent-net.rules` file) should be enough to let you to run the new scheme, but on Debian 8 "Jessie" you'll need to actively set it to `net.ifnames=1`. /* And on Debian 7 "Wheezy" it probably just plain won't work, but should you really be connecting that to the Internet anyway? */

<<Anchor(trigger)>>
 * TRIGGERING THE SWITCH:: the brave way of finding out what network interface names you'll get without a `/etc/udev/rules.d/70-persistent-net.rules` file is to delete it (and update your [[#initrd|initrd]] before you reboot), but you don't need to go that far. Just renaming it (e.g. to `70-persistent-net.rules.old`) or commenting out particular lines should be enough. See the `udev` `README.Debian.gz` file. Note that it is possible to have a mixed system with (say) an `enp1s1` named from its hardware path alongside a `wlan0` still defined as a "persistent" name.

 * THE ID_NET_NAME_ HIERARCHY:: if you're ignoring ID_NET_NAME_SOMETHING on the assumption that anything you don't understand probably isn't important, you need to reread [[#hierarchy|the above]] - the general rule is, if you don't recognise it, it'll mess things up.

 * PREFIXES AND PATHS:: wired devices get a prefixed `en-` for Ethernet, wireless ones get `wl-` (and there are also a few more obscure possibilities such as `ib-` for !InfiniBand); then in principle it's possible to decipher all the following [[https://www.freedesktop.org/software/systemd/man/systemd.net-naming-scheme.html|sequences of code letters plus hex digits]] that encode hardware topology. But there's not much point trying to learn all the details, since the only workable way of predicting what ID_NET_NAMEs an interface will get is to ask `udevadm`, which will tell you the full strings.

 * MYSTERY PRIORITIES:: if you look at `/lib/systemd/network/99-default.link` (at least on Debian 10 "Buster") /* (Stretch didn't have `keep`, but treated it as present by default) */ the standard hierarchy of priorities is "`keep kernel database onboard slot path`" (note that `mac` only gets considered via a different mechanism). Those first three are poorly documented; the docs confirm that they're real, but the definitions given aren't much help. It would make sense if `keep` meant "if a custom rule gets there first then `udev` won't re-rename the interface"; `kernel` seems to mean something similar for interfaces that the kernel says not to re-rename, but when this happens is unclear. /* One plausible guess is that it means "net.ifnames=0, so do nothing"; but see [[https://salsa.debian.org/systemd-team/systemd/commit/40df7df1f6c4812fa7b074c5d48dcfce6ce8a782]] - baffling. */

 * USB DEVICES:: since they might get plugged into a different socket each time, these use ID_NET_NAME_MAC - automated via `/lib/udev/rules.d/73-usb-net-by-mac.rules`.

 * VIRTUAL MACHINES:: on virtual machines (according to the `udev` README) you will need to remove the files `/etc/systemd/network/99-default.link` and (if using virtio network devices) `/etc/systemd/network/50-virtio-kernel-names.link`, then [[#initrd|rebuild the initrd]].
Line 246: Line 173:
 * INITRD SKEW:: it's all very well having everything sorted out in
`/etc`; but to make sure your initramfs doesn't contain out-of-date
versions of important `systemd` unit files, regenerate it with

{{{
        sudo update-initramfs -u
}}}

 * LEFTOVERS:: the old system first started being publicly deprecated
in `NEWS` files back in mid-2015, so this upgrade has been hanging
ominously over people's heads for quite long enough now that you're
entitled to forgive yourself if you've forgotten something you did
last time the subject came up. Did you set up a `net.ifnames=0` kernel
parameter, and/or mask some `systemd` config file? Check your
administrative logbooks. What do you mean, you don't keep logs?

 * USB DEVICES:: since they might get plugged into a different socket
each time, these use ID_NET_NAME_MAC - automated via
`/lib/udev/rules.d/73-usb-net-by-mac.rules`.

 * VIRTUAL MACHINES:: on virtual machines (according to the `udev`
README) you will need to remove the files
`/etc/systemd/network/99-default.link` and (if using virtio network
devices) `/etc/systemd/network/50-virtio-kernel-names.link`,
then [[#initrd|rebuild the initrd]].

 * UNPREDICTABILITY:: it turns out even after all this there are still
reported cases of interfaces changing their name on a reboot. All that
needs to happen is that some buggy BIOS, or some new, less buggy
version of a driver module, changes its mind about whether or not your
hardware counts as the kind that should have an ONBOARD or SLOT name.

<<Anchor(alt)>>
= THE ALTERNATIVE APPROACH =

You can also use `.link` files to set up naming schemes to suit
yourself - so for instance if you have two PCs each of which has only
a single wireless card, but one calls it `wlp0s1` and the other
`wlp1s0`, you can arrange for them both to use the name `wifi0` to
simplify sharing firewall configurations. For details see
[[DebianMan:5/systemd.link|systemd.link(5)]].

It's also possible to reorganize the naming policy by overriding the
!NamePolicy defined in `/lib/systemd/network/99-default.link` (so that
for instance all network cards are named by MAC address).
 * INITRD SKEW:: it's all very well having everything sorted out in `/etc`, but interface renaming has to happen very early during boot; to make sure your initrd doesn't contain out-of-date versions of important systemd files, regenerate it with `sudo update-initramfs -u` /* Unless of course you're running without an initrd, in which case presumably you'll know what to do */

 * LEFTOVERS:: the old persistent-names system first started being publicly deprecated in `NEWS` files back in mid-2015, so this upgrade has been hanging ominously over people's heads for a long time. Are you sure you didn't do something about it the last time the subject came up, like setting up a `net.ifnames=0` kernel parameter, and/or masking some systemd config file? If so, this may result in confusing symptoms when you try to go over to the new system. Check your administrative logbooks. What do you mean, you don't keep logs?

 * TO BE DETERMINED:: if you're aware of extra sources of complications not accounted for here involving (for instance) non-systemd initsystems; minor ports; `systemd-networkd`; or something else that has turned up since this was first written, please add them here.

 * UNPREDICTABILITY:: it turns out even after all this there are still reported cases of interfaces changing their name on a reboot. All that needs to happen is that some buggy BIOS (or some new, less buggy version of a driver module, or systemd's naming policy) changes its mind about some detail like whether or not your hardware counts as the kind that should have an ONBOARD name. There are even [[https://utcc.utoronto.ca/~cks/space/blog/linux/PCINamesNotStable|reports]] of devices changing their PCI-port numbering due to other hardware being installed.

<<Anchor(nets)>>
== Safety nets ==

 * DEAD-MAN'S-HANDLE:: if you're rebooting with `70-persistent-net.rules` [[#trigger|renamed as]] `70-persistent-net.rules.old`, and there's a danger that you might find yourself locked out, you can set up a precautionary script (called by `/etc/rc.local`, @reboot cronjob, or systemd timer) that waits a few minutes, then copies the file back again, rebuilds the initrd, and reboots. Extra fancy checks and beeping noises and so on are possible, but increase the risk of failure.

 * HOKEY-COKEY MIGRATION:: if there are two interfaces, either of which could be used to give you remote access, both [[#legacy|currently named via]] `70-persistent-net.rules`, you can comment one of them out of the file, [[#initrd| update the initrd]], reboot to see what it's called now, then go back and reactivate it before commenting out the other. If everything's still going okay, you can finish it all off. /* Don't start drinking until at least stage three. */

 * CATCHALL INTERFACE NAMING:: if your SSH interface is expected to come back as `enp0s1` after the reboot, and that's what you've got configured in `/etc/network/interfaces`, but instead it decides to call itself `eno0`, that's a problem - but one that a sufficiently cautious admin can guard against by having entire duplicate stanzas in the `interfaces` file to define the same IP address for ''every'' name it might plausibly come back with, including `eno0`, `ens0`, `eth0`, and so on. Mind you, this still won't help if it comes back as `enp7s1`.

 * MANUAL OVERRIDE:: one way of being sure is to avoid trusting `udev` to make its own mind up about what your crucial network interface should be called; switch it over to a name [[#custom|defined]] in a custom `.link` file.

 * EXTRA SUGGESTIONS GO HERE:: ...

----

<<Anchor(custom)>>
= CUSTOM SCHEMES USING .LINK FILES =

The scheme detailed above is the new standard default, but there's also a canonical way of overriding the default: you can use `.link` files to set up naming policies to suit your needs. Thus for instance if you have two PCs each of which has only a single wireless card, but one calls it `wlp0s1` and the other `wlp1s0`, you can arrange for them both to use the name `wifi0` to simplify sharing firewall configurations. For details see [[DebianMan:5/systemd.link|systemd.link(5)]].

Here's a relatively futureproof "manual" version of the example [[#legacy|given above]]: /* It Works For Me, at least with corrected MAC */
{{{
 #/etc/systemd/network/10-persistent-net.link
 [Match]
 MACAddress=01:23:45:67:89:ab

 [Link]
 Name=eth0
}}}

It is also possible to reorganize the naming policy by overriding `/lib/systemd/network/99-default.link`, for instance to insist that all network interfaces are named purely by MAC address: /* It Works For Me, but I prefer the above */

{{{
 #/etc/systemd/network/99-default.link
 [Link]
 NamePolicy=mac
 MACAddressPolicy=persistent
}}}

----
Line 295: Line 225:
[[https://www.debian.org/releases/stretch/amd64/release-notes/ch-whats-new.en.html#new-interface-names|Stretch
Release Notes]]

[[https://www.debian.org/releases/buster/amd64/release-notes/ch-information.en.html#migrate-interface-names|Buster
Release Notes]]

There's some information in `/usr/share/doc/udev/README.Debian.gz`,
and more in

[[https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/]]

The big problem with the latter is that it delegates all its technical
details to a link pointing at the sourcecode:
[[https://github.com/systemd/systemd/blob/master/src/udev/udev-builtin-net_id.c]]
...but unfortunately most of the useful comments that used to be at
the top of that file ha
ve been thrown out, so you need to find your
way back through the `git` tree to a pre
vious version such as
[[https://github.com/systemd/systemd/blob/eefe36e64c1a583bb9470884ed92115e0ce4647e/src/udev/udev-builtin-net_id.c]]

A guide that mentions ID_NET_NAME_FROM_DATABASE:
[[https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/]]

General guides to overriding [[Systemd]] configuration:
[[https://askubuntu.com/questions/659267/how-do-i-override-or-configure-systemd-services]]
[[https://wiki.archlinux.org/index.php/systemd]]

----
CategoryNetwork
[[https://www.debian.org/releases/stretch/amd64/release-notes/ch-whats-new.en.html#new-interface-names|Stretch]] and [[https://www.debian.org/releases/buster/amd64/release-notes/ch-information.en.html#migrate-interface-names|Buster]] Release Notes

The nearest the upstream docs ever got to a canonical migration-HOWTO was
[[https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/]]. The big problem with this was that it delegated all its technical details to a link pointing at the sourcecode:[[https://github.com/systemd/systemd/blob/master/src/udev/udev-builtin-net_id.c]] ...but most of the useful comments that used to be at the top of that file were then thrown out, so you need to find your way back through the `git` tree to a previous version such as [[https://github.com/systemd/systemd/blob/eefe36e64c1a583bb9470884ed92115e0ce4647e/src/udev/udev-builtin-net_id.c]]. Meanwhile, the page now claims to be obsolete, and points instead to [[https://www.freedesktop.org/software/systemd/man/systemd.net-naming-scheme.html]], which is much less helpful. So the nearest thing left to an official HOWTO is probably `/usr/share/doc/udev/README.Debian.gz` (though it doesn't cover the "how to predict the names" part at all).

A guide that mentions ID_NET_NAME_FROM_DATABASE:[[https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/]]

General guides to overriding [[systemd]] configuration:
[[https://askubuntu.com/questions/659267/how-do-i-override-or-configure-systemd-services]], [[https://wiki.archlinux.org/index.php/systemd]] /* This is another topic that's enough of an FAQ that I was rather expecting there to be an official upstream HOWTO, but apparently not */

----
CategoryNetwork, CategorySystemAdministration

'''Keywords''': persistent, predictable, NIC, wlan, eth, migrate

See below for a "Predictable Names" Migration HOWTO.

Anything that changes the names of your network interfaces may result in the machine suddenly not being reachable over SSH, so if you're editing settings on a remote server, plan your changes carefully and doublecheck your safety nets.

This page deals with the various schemes by which wired and wireless network interfaces are assigned names - that is, the underlying system labels like eth0 or wlx800e1319c734. It has nothing to do with the "connection profile" names used by apps such as NetworkManager, like "Wired connection 1".


THE ORIGINAL SIMPLE SCHEME

Back in the nineties, eth0, eth1, etc were simply assigned by the kernel.

Why it was abandoned

At least in theory, if module probes completed in a different order, eth0 and eth1 might switch places on successive boots. As boot processes became less linear and interfaces became more hotpluggable this became more of a concern.

How to get it back

If you wipe out all other name-assignment mechanisms then you'll be left with this one.

The simple way of disabling the whole current interface naming scheme (which you might want to try for one-off testing) is just to boot with the kernel parameter net.ifnames=0, which can be set in an interactive grub session at boot or made persistent by editing /etc/default/grub and running update-grub.

Alternatively, you can override /lib/systemd/network/99-default.link, with a custom version in /etc/systemd/network/, or similarly override /lib/udev/rules.d/80-net-setup-link.rules, or mask the latter by using a /dev/null symlink instead of a custom version, or... there seem to be lots of ways of doing this, so make sure you haven't done it in more than one way or it'll trip you up in a couple of years when you try to undo it. See the external links below on standard methods for overriding systemd configuration. Oh, and beware of initrd skew.

(If you've still got a working legacy 70-persistent-net.rules file, the net.ifnames=0 flag doesn't deactivate that, so you'd need to see the instructions below for temporarily disabling that file. It's probably also possible to do this by masking enough of systemd.)


THE "PERSISTENT NAMES" SCHEME

This scheme, introduced somewhere around Debian 5 "Lenny", used udev to identify interfaces by MAC address and assign a fixed interface number to any interface it recognized (writing the rules to /etc/udev/rules.d/70-persistent-net.rules). This could have annoying side-effects (e.g. if you were replacing a machine's sole NIC, you'd also have to take special care to ensure it took over as network interface number 0) but these were minor and predictable.

A sample .rules entry:

 #/etc/udev/rules.d/70-persistent-net.rules
 # PCI device 0x8086:0x100e (e1000e)
 SUBSYSTEM=="net", ACTION=="add", DRIVERS=="?*", ATTR{address}=="01:23:45:67:89:ab", ATTR{type}=="1", KERNEL=="eth*", NAME="eth0"

Since Debian 9 "Stretch", newly installed machines no longer start with an /etc/udev/rules.d/70-persistent-net.rules file, though such files are maintained if they still exist (with new lines added for newly installed network hardware). Debian 10 "Buster" additionally lacks the /lib/udev/rules.d/75-persistent-net-generator.rules file that appends to it, though legacy 70-persistent-net.rules files are still honored. This support is expected to be removed from Debian 11 "Bullseye" at some time before it is released.

Why this one was abandoned

This still had subtle race conditions, required /etc to be on a writable file system, and had problems with virtualization, so it's no longer supported upstream; the plan (still taken for granted in most of the documentation) was for it not to be supported in Debian 10 "Buster", but in the end it has scraped through into another release cycle.

How to cling to it for now

If you've got a working "legacy" /etc/udev/rules.d/70-persistent-net.rules file and want to stick with it, you can safely upgrade through Debian 9 "Stretch" and Debian 10 "Buster"; udev on these releases still respects that file if present (and will accept a freshly created one). However, bear in mind that on Buster you'll need to maintain its contents yourself, and eventually even this legacy support will end, so you should be ready to switch to a different scheme before that happens.

How to let go and move on

If you're currently running something newer than Debian 8 "Jessie" with a legacy 70-persistent-net.rules file but have decided to switch to the new regime, you can do that just by disabling the .rules file (then updating the initrd before you reboot); see the udev README.Debian.gz and the more detailed guide below.


MISCELLANEOUS OLD SCHEMES

(Contributions welcome)

Several workarounds for renaming interfaces grew up in the early days of hotpluggable wireless interfaces, but if they still work it'll be because like ifrename they now use udev rules under the hood. It's not clear what remaining advantage this has over the canonical .link approach - is it perhaps useful for non-systemd machines?

Old releases of RedHat (among others) used a "biosdevname" system, but that's never been supported under Debian. If you need to know about it there's bound to be documentation somewhere.


THE "PREDICTABLE NAMES" SCHEME

The new scheme uses names usually derived from the location of the interface in terms of hardware buses etc: eno1, wlp1s3. The idea was that this provides "Predictable Names", though as it turns out the main thing that's predictable about it is that calling it this will cause furious users to pop up disputing the appropriateness of that name. (Can we just skip all that here, please?)

How to cope with it on fresh installs

This should be easy enough; before you start configuring firewalls etc., just look at (e.g.) the output of ip a and note the names of the interfaces. Unlike the old days, when the only way to guess which cable was plugged into eth0 and which was eth1 was to keep track of MAC addresses, this system provides extra clues in the interface names.

How to migrate to this scheme on upgraded systems

It's advisable to do this as a separate migration in its own right, not as part of a general distribution upgrade. However, if your PC only has one network interface and not much is at stake you can try:

Strategy A

  • wait until udev breaks your networking, if it's going to, or trigger the change yourself.

  • ask ip what new name it's using, and fix your configuration files.

You should probably at least check in advance to see what files hard-code interface names, by running something like

        sudo rgrep wlan0 /etc

Obvious likely hits include /etc/network/interfaces and configuration files for firewalls, wifi, DHCP... but it's possible that (e.g.) a laptop with a single wifi interface managed by NetworkManager might need no fixing at all.

Strategy B

This strategy, more or less compulsory for remote servers, runs along the lines of:

  • devise a migration plan
  • consider a dead-man's-handle etckeeper cronjob or the like

  • consult the list of corner cases below and search the Internet for others

  • once you're sure you're safe, implement your migration plan
  • you've got backups, right?

To find out what names udev would be choosing between if you switched over to the new system, first get a list of the network devices the system knows about:

        echo /sys/class/net/*

For each device path (other than /sys/class/net/lo), ask udevadm what NET_IDs it knows:

        udevadm test-builtin net_id /sys/class/net/enp0s1 2>/dev/null

It's likely to tell you about things like ID_OUI_FROM_DATABASE and an ID_NET_NAMING_SCHEME, but the lines that matter are the ones (given in random order) starting with ID_NET_NAME_. One of these is the name that udev will give priority to - the list of candidates may be so short that all you need to know is that ..._PATH beats ..._MAC, but there are also some rarer possibilities, and in general if something unusual shows up then it will take priority.

From highest priority to lowest, the list is:

  1. ID_NET_NAME_FROM_DATABASE= Very rare and not to be confused with ID_OUI_FROM_DATABASE; if present, it outranks any other ID_NET_NAME. The database is hardcoded into udev and has only one known entry, the spooky-sounding idrac.

  2. ID_NET_NAME_ONBOARD= Appears for some but not all kinds of onboard network card - it's usually a nice simple name like eno0 or wlo0.

  3. ID_NET_NAME_SLOT= Appears for some PCI-hotplug cards. Usually looks like ens0 or wls0. (Does this ever occur alongside _ONBOARD?)

  4. ID_NET_NAME_PATH= Always present; usually something just complicated enough to be easy to forget, like wlp3s5 or enp1s3f0. Note that all numbers are in hex.

  5. ID_NET_NAME_MAC= Also always present, but with a low enough priority that by default it won't be used; e.g. wlx800e1319c734

One good reason for doing the migration separately from a dist-upgrade is that the answers you get from (e.g.) Stretch's udevadm aren't absolutely guaranteed to be identical to the answers you get from Buster's. This further implies that any time you update systemd and reboot there's a chance your server might fall off the net, which seems like a good argument for using a customized scheme at least for the interface you're SSHing in on.

Complications and corner cases

(Additions welcome, but please try to avoid ballooning this section with tales of "I don't know how this happened but it all went wrong for me"...)

* ANTIQUE SYSTEMS

on Debian 9 "Stretch" or newer, merely booting without a net.ifnames=0 override (and without a 70-persistent-net.rules file) should be enough to let you to run the new scheme, but on Debian 8 "Jessie" you'll need to actively set it to net.ifnames=1.

* TRIGGERING THE SWITCH

the brave way of finding out what network interface names you'll get without a /etc/udev/rules.d/70-persistent-net.rules file is to delete it (and update your initrd before you reboot), but you don't need to go that far. Just renaming it (e.g. to 70-persistent-net.rules.old) or commenting out particular lines should be enough. See the udev README.Debian.gz file. Note that it is possible to have a mixed system with (say) an enp1s1 named from its hardware path alongside a wlan0 still defined as a "persistent" name.

* THE ID_NET_NAME_ HIERARCHY

if you're ignoring ID_NET_NAME_SOMETHING on the assumption that anything you don't understand probably isn't important, you need to reread the above - the general rule is, if you don't recognise it, it'll mess things up.

* PREFIXES AND PATHS

wired devices get a prefixed en- for Ethernet, wireless ones get wl- (and there are also a few more obscure possibilities such as ib- for InfiniBand); then in principle it's possible to decipher all the following sequences of code letters plus hex digits that encode hardware topology. But there's not much point trying to learn all the details, since the only workable way of predicting what ID_NET_NAMEs an interface will get is to ask udevadm, which will tell you the full strings.

* MYSTERY PRIORITIES

if you look at /lib/systemd/network/99-default.link (at least on Debian 10 "Buster") the standard hierarchy of priorities is "keep kernel database onboard slot path" (note that mac only gets considered via a different mechanism). Those first three are poorly documented; the docs confirm that they're real, but the definitions given aren't much help. It would make sense if keep meant "if a custom rule gets there first then udev won't re-rename the interface"; kernel seems to mean something similar for interfaces that the kernel says not to re-rename, but when this happens is unclear.

* USB DEVICES

since they might get plugged into a different socket each time, these use ID_NET_NAME_MAC - automated via /lib/udev/rules.d/73-usb-net-by-mac.rules.

* VIRTUAL MACHINES

on virtual machines (according to the udev README) you will need to remove the files /etc/systemd/network/99-default.link and (if using virtio network devices) /etc/systemd/network/50-virtio-kernel-names.link, then rebuild the initrd.

* INITRD SKEW

it's all very well having everything sorted out in /etc, but interface renaming has to happen very early during boot; to make sure your initrd doesn't contain out-of-date versions of important systemd files, regenerate it with sudo update-initramfs -u

* LEFTOVERS

the old persistent-names system first started being publicly deprecated in NEWS files back in mid-2015, so this upgrade has been hanging ominously over people's heads for a long time. Are you sure you didn't do something about it the last time the subject came up, like setting up a net.ifnames=0 kernel parameter, and/or masking some systemd config file? If so, this may result in confusing symptoms when you try to go over to the new system. Check your administrative logbooks. What do you mean, you don't keep logs?

* TO BE DETERMINED

if you're aware of extra sources of complications not accounted for here involving (for instance) non-systemd initsystems; minor ports; systemd-networkd; or something else that has turned up since this was first written, please add them here.

* UNPREDICTABILITY

it turns out even after all this there are still reported cases of interfaces changing their name on a reboot. All that needs to happen is that some buggy BIOS (or some new, less buggy version of a driver module, or systemd's naming policy) changes its mind about some detail like whether or not your hardware counts as the kind that should have an ONBOARD name. There are even reports of devices changing their PCI-port numbering due to other hardware being installed.

Safety nets

* DEAD-MAN'S-HANDLE

if you're rebooting with 70-persistent-net.rules renamed as 70-persistent-net.rules.old, and there's a danger that you might find yourself locked out, you can set up a precautionary script (called by /etc/rc.local, @reboot cronjob, or systemd timer) that waits a few minutes, then copies the file back again, rebuilds the initrd, and reboots. Extra fancy checks and beeping noises and so on are possible, but increase the risk of failure.

* HOKEY-COKEY MIGRATION

if there are two interfaces, either of which could be used to give you remote access, both currently named via 70-persistent-net.rules, you can comment one of them out of the file, update the initrd, reboot to see what it's called now, then go back and reactivate it before commenting out the other. If everything's still going okay, you can finish it all off.

* CATCHALL INTERFACE NAMING

if your SSH interface is expected to come back as enp0s1 after the reboot, and that's what you've got configured in /etc/network/interfaces, but instead it decides to call itself eno0, that's a problem - but one that a sufficiently cautious admin can guard against by having entire duplicate stanzas in the interfaces file to define the same IP address for every name it might plausibly come back with, including eno0, ens0, eth0, and so on. Mind you, this still won't help if it comes back as enp7s1.

* MANUAL OVERRIDE

one way of being sure is to avoid trusting udev to make its own mind up about what your crucial network interface should be called; switch it over to a name defined in a custom .link file.

* EXTRA SUGGESTIONS GO HERE
...


CUSTOM SCHEMES USING .LINK FILES

The scheme detailed above is the new standard default, but there's also a canonical way of overriding the default: you can use .link files to set up naming policies to suit your needs. Thus for instance if you have two PCs each of which has only a single wireless card, but one calls it wlp0s1 and the other wlp1s0, you can arrange for them both to use the name wifi0 to simplify sharing firewall configurations. For details see systemd.link(5).

Here's a relatively futureproof "manual" version of the example given above:

 #/etc/systemd/network/10-persistent-net.link
 [Match]
 MACAddress=01:23:45:67:89:ab

 [Link]
 Name=eth0

It is also possible to reorganize the naming policy by overriding /lib/systemd/network/99-default.link, for instance to insist that all network interfaces are named purely by MAC address:

 #/etc/systemd/network/99-default.link
 [Link]
 NamePolicy=mac
 MACAddressPolicy=persistent


EXTERNAL REFERENCES

Stretch and Buster Release Notes

The nearest the upstream docs ever got to a canonical migration-HOWTO was https://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames/. The big problem with this was that it delegated all its technical details to a link pointing at the sourcecode:https://github.com/systemd/systemd/blob/master/src/udev/udev-builtin-net_id.c ...but most of the useful comments that used to be at the top of that file were then thrown out, so you need to find your way back through the git tree to a previous version such as https://github.com/systemd/systemd/blob/eefe36e64c1a583bb9470884ed92115e0ce4647e/src/udev/udev-builtin-net_id.c. Meanwhile, the page now claims to be obsolete, and points instead to https://www.freedesktop.org/software/systemd/man/systemd.net-naming-scheme.html, which is much less helpful. So the nearest thing left to an official HOWTO is probably /usr/share/doc/udev/README.Debian.gz (though it doesn't cover the "how to predict the names" part at all).

A guide that mentions ID_NET_NAME_FROM_DATABASE:https://major.io/2015/08/21/understanding-systemds-predictable-network-device-names/

General guides to overriding systemd configuration: https://askubuntu.com/questions/659267/how-do-i-override-or-configure-systemd-services, https://wiki.archlinux.org/index.php/systemd


CategoryNetwork, CategorySystemAdministration

Keywords: persistent, predictable, NIC, wlan, eth, migrate