[arch-releng] Systemd related issues with 2014.09.01 ISO image
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
Greetings,
Pierre
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I can't move 216-2 to [core] because lz4 compression apparently doesn't work at all. I'm going to have a 216-3 in [testing] today which reverts the lz4 support.
d
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:22:
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I thought this should be really unique... Are there any use cases where this can make problems? I do boot several instances of the same ISO in one network.
On Mon, Sep 01, 2014 at 07:36:16PM +0200, Christian Hesse wrote:
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:22:
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I thought this should be really unique... Are there any use cases where this can make problems? I do boot several instances of the same ISO in one network.
I don't think this would ever be a problem in normal usage of a live environment. However, it'd be pretty simple to randomize the UUID -- generate one via method of your choosing (there's at least 5 methods available) as a late_hook().
d
Am 01.09.2014 um 19:51 schrieb Dave Reisner:
On Mon, Sep 01, 2014 at 07:36:16PM +0200, Christian Hesse wrote:
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:22:
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I thought this should be really unique... Are there any use cases where this can make problems? I do boot several instances of the same ISO in one network.
I don't think this would ever be a problem in normal usage of a live environment. However, it'd be pretty simple to randomize the UUID -- generate one via method of your choosing (there's at least 5 methods available) as a late_hook().
Yes, IMO we should do that.
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:51:
On Mon, Sep 01, 2014 at 07:36:16PM +0200, Christian Hesse wrote:
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:22:
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I thought this should be really unique... Are there any use cases where this can make problems? I do boot several instances of the same ISO in one network.
I don't think this would ever be a problem in normal usage of a live environment. However, it'd be pretty simple to randomize the UUID -- generate one via method of your choosing (there's at least 5 methods available) as a late_hook().
I already do. ;)
I stumbled on this issue earlier today and fixed it. That's why I could answer that fast.
On Mon, Sep 1, 2014 at 7:36 PM, Christian Hesse list@eworm.de wrote:
Dave Reisner d@falconindy.com on Mon, 2014/09/01 13:22:
On Mon, Sep 01, 2014 at 07:14:24PM +0200, Christian Hesse wrote:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
Not sure what you mean by this -- systemd has always generated a machine-id on first install, making live CDs all have the same machine ID. This isn't a problem. The newly installed system will always have a separate and unique machine-id of its own..
I thought this should be really unique... Are there any use cases where this can make problems? I do boot several instances of the same ISO in one network.
Yeah, the machine-id should really be unique. For instance networkd will use this whenever a unique, persistent per-machine value is needed, for instance to generate MAC addresses or DHCP identifiers (I assume there are also other such users).
On a live system, you could simply ship an empty /etc/machine-id and PID1 will generate a random one in /run and overmount that file into /etc.
Cheers,
Tom
Am 01.09.2014 um 19:14 schrieb Christian Hesse:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
In that case, you can probably get back to the old state by masking systemd-firstboot.service.
Am 01.09.2014 um 19:22 schrieb Thomas Bächler:
Am 01.09.2014 um 19:14 schrieb Christian Hesse:
Pierre Schmitz pierre@archlinux.de on Mon, 2014/09/01 19:01:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
This is a known issue with systemd 216-1:
https://bugs.archlinux.org/task/41711
Though I doubt the fix from systemd 216-2 is a good idea for live media... We want a new machine-id for every system, no?
In that case, you can probably get back to the old state by masking systemd-firstboot.service.
I spoke too soon, sorry. Let's simply generate a machine-id for the ISO instead.
Am 01.09.2014 um 19:01 schrieb Pierre Schmitz:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
Greetings,
Pierre
I just pushed archiso v19 which fixes this. Additionally, it ensures that each live system has a unique machine id.
Systemd 216-3 sort of works around this as well, but v19 solves this properly.
Am 02.09.2014 20:42, schrieb Thomas Bächler:
Am 01.09.2014 um 19:01 schrieb Pierre Schmitz:
Hi,
I just build the new ISO image. But I noticed some strange things are going on with the systemd services. It seems all available services are enabled and get started. Of course some of them will then fail: https://paste.archlinux.de/Blu/
I did not look into it yet; cant tell when though.
Greetings,
Pierre
I just pushed archiso v19 which fixes this. Additionally, it ensures that each live system has a unique machine id.
Systemd 216-3 sort of works around this as well, but v19 solves this properly.
Thanks; kind of a weird feature. I'll uploada new ISO soon; probably tomorrow.
Greetings,
Pierre
participants (5)
-
Christian Hesse
-
Dave Reisner
-
Pierre Schmitz
-
Thomas Bächler
-
Tom Gundersen