Bug#770776: release-notes: Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
Control: tags -1 pending On 2014-12-30 16:59, Tomas Pospisek wrote: On Sun, 7 Dec 2014, Adam D. Barratt wrote: [...] So please check, if required criticise and please apply the patch. Applied with modifications (please see attached patch). The summary: * Some minor wording changes / sentence movements. * Move section to issues rather than upgrading * (More) Cross-linking between LXC section and systemd migration section. * Add a note that the LXC section does *not* apply to normal end-user systems. Thanks you all for contributing and considering and have a wonderful 2015! *t PS: Many thanks to Niels and Holger for their kind and supportive words! This patch is dedicated to you guys!! Same to you and you are welcome :) PPS: Adam and Niels and Daniel, could you please have a look at #774204. If you don't object Daniel, I'd propose to NMU that and ask you release team to have it included in jessie. Please follow up in #774204! Ok, will follow up. ~Niels From 8c90bfcdccee633689895828084637cf5cb2e682 Mon Sep 17 00:00:00 2001 From: nthykier nthykier@313b444b-1b9f-4f58-a734-7bb04f332e8d Date: Thu, 1 Jan 2015 12:20:39 + Subject: [PATCH] en/issues: Document LXC migration issues Amendments by nthykier: * Some minor wording changes / sentence movements. * Move section to issues rather than upgrading * (More) Cross-linking between LXC section and systemd migration section. * Add a note that the LXC section does *not* apply to normal end-user systems. Signed-off-by: Niels Thykier ni...@thykier.net git-svn-id: svn+ssh://svn.debian.org/svn/ddp/manuals/trunk/release-notes@10573 313b444b-1b9f-4f58-a734-7bb04f332e8d --- en/issues.dbk | 83 +++ 1 file changed, 78 insertions(+), 5 deletions(-) diff --git a/en/issues.dbk b/en/issues.dbk index b9dcf29..fd35ba7 100644 --- a/en/issues.dbk +++ b/en/issues.dbk @@ -210,11 +210,15 @@ $ echo 'openssh-server openssh-server/permit-root-login boolean true' | debconf- preference for another init such as systemitem role=packagesysvinit-core/systemitem or systemitem role=packageupstart/systemitem, it is recommended to setup -APT pinning prior to the upgrade. As an example, to prevent -systemitem role=packagesystemd/systemitem from being -installed during the upgrade, you can create a file called -filename/etc/apt/preferences.d/local-pin-init/filename with -the following contents: +APT pinning prior to the upgrade. This may also be required if +you are upgrading LXC containers before the host. In this case, +please refer to xref linkend=issues-lxc-wheezy-host / + /para + paraAs an example, to prevent systemitem +role=packagesystemd/systemitem from being installed during the +upgrade, you can create a file called +filename/etc/apt/preferences.d/local-pin-init/filename with the +following contents: /para screen Package: systemd-sysv @@ -300,6 +304,75 @@ ProtectSystem=yes /para /section +section id=issues-lxc +!-- Wheezy to Jessie -- +titleUpgrade considerations for LXC hosts and containers/title + +note + para +This section only applies to systems that has LXC containers and +hosts. Normal end user systems usually do not have these. + /para +/note +para +The upgrade from wheezy to jessie will migrate your system to the +systemd init system by default (see xref +linkend=systemd-upgrade-default-init-system /). +/para +para +When upgrading a LXC container, respectively a LXC virtual +machine, this will have different consequences depending on +whether the emphasishost system/emphasis has already +been upgraded to jessie or not. +/para + +section id=issues-lxc-wheezy-host +titleUpgrading LXC guests running on wheezy hosts/title + +para + If you are upgrading a LXC guest container that is running on a + emphasisWheezy host/emphasis system, then you will need to + prevent the guest from being automatically migrated to systemd. You + prevent the migration via pinning, as described in xref + linkend=systemd-upgrade-default-init-system /. +/para +para + This is required as the Wheezy host lacks functionality to boot a + system running systemd. +/para +para + You should be able to switch over to systemd inside the LXC guest + once you have upgraded the emphasishost/emphasis system to + Jessie. See the next paragraph for things that need to be adapted on + Jessie hosts. +/para +/section + +section id=issues-lxc-jessie-host +titleUpgrading LXC guests running on jessie hosts/title + +para + In order to be able to boot LXC guests with systemd, you need to + adapt your LXC container configuration. The container configuration + can usually be found in + filename/var/lib/lxc/replaceableCONTAINER_NAME/replaceable/config/filename + You need to add the following two settings to the configuration: +/para +screen +lxc.autodev = 1 +lxc.kmsg = 0
Bug#770776: release-notes: Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On Sun, 7 Dec 2014, Adam D. Barratt wrote: Is this still relevant after the upload of lxc 1.0.6-5? Yes, it still is. There are still manual steps needed during upgrade. Attached you'll find my second try at a patch to the release notes. If my suggested addition is acceptable and not too far off the mark, then I suggest to apply it *now*. Changes and improvements can be incorporated into the patched release-notes. The reason is, that it took me one month to come back to this issue and I don't know when I'll have the next opportunity to work on this. And I'd rather have the problem documented in the interest of the users. That said, please do criticise, if there's something to improve. So please check, if required criticise and please apply the patch. Thanks you all for contributing and considering and have a wonderful 2015! *t PS: Many thanks to Niels and Holger for their kind and supportive words! This patch is dedicated to you guys!! PPS: Adam and Niels and Daniel, could you please have a look at #774204. If you don't object Daniel, I'd propose to NMU that and ask you release team to have it included in jessie. Please follow up in #774204!diff --git en/upgrading.dbk en/upgrading.dbk index 13b1a30..f7dc09c 100644 --- en/upgrading.dbk +++ en/upgrading.dbk @@ -1136,6 +1136,79 @@ entries. You should use commandvisudo/command for this: /screen /section +section id=issues-lxc +titleLXC/title + +para +The upgrade from wheezy to jessie will migrate your system to +the systemd init system. +/para +para +When upgrading a LXC container, respectively a LXC virtual +machine, this will have different consequences depending on +whether the emphasishost system/emphasis has allready +been upgraded to jessie or not. +/para + +section id=issues-lxc-wheezy-host +titleUpgrading LXC guests running on wheezy hosts/title + +para +If you are upgrading a LXC guest container that is running on a +emphasiswheezy host/emphasis system, then you will need to +prevent the guest from being automatically migrated to systemd. +/para +para +If you don't prevent the system from migrating to systemd, then +you will risk, that your guest system will not be able to boot. +/para +para +This is due to functionality missing in wheezy hosts, that are +required to boot a jessie system with systemd. +/para +para +You prevent the migration via pinning, as described in chapter +xref linked=#systemd-upgrade-default-init-system/. +/para +para +You should be able to switch over to systemd inside the LXC guest +once you have upgraded the emphasishost/emphasis system to jessie. See the next +paragraph for things that need to be adapted on jessie hosts. +/para +/section + +section id=issues-lxc-jessie-host +titleUpgrading LXC guests running on jessie hosts/title + +para +In order to be able to boot LXC guests with systemd, you need +to adapt your LXC container configuration. +/para +para +The container configuration can usually be found in +/para +screen +/var/lib/lxc/replaceableCONTAINER_NAME/replaceable/config +/screen +para +You need to add the following two setting to the configuration: +/para +screen +lxc.autodev = 1 +lxc.kmsg = 0 +/screen +/section + +section id=issues-lxc-further-docu +titleFurther information/title +para +You can find further information on LXC in Debian +ulink url=url-wiki;LXCin the Debian wiki/ulink. +/para +/section + +/section + /section !-- End of 'trouble' section -- /section
Bug#766233: Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On Sun, 7 Dec 2014, Adam D. Barratt wrote: On Mon, 2014-11-24 at 00:19 +0100, Tomas Pospisek wrote: LXC VMs become unbootable when upgraded to jessie. The upgrade migrates systems automatically to systemd, after which the VMs won't be able to boot. Is this still relevant after the upload of lxc 1.0.6-5? Yes it still is. From my terminal, fresh installation of a wheezy VM on a wheezy host, upgrading the VM to jessie: root@host ~ # lxc-start -n tpotest2 lxc-start: unknow key lxc.autodev lxc-start: failed to read configuration file The problem is still, that a wheezy *host* doesn't have the required features to run a jessie *guest* with systemd. *t PS: I was a bit off-Debian for the last month or so, thanks for pinging this Adam! -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#766216: Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On Mon, 2014-11-24 at 00:19 +0100, Tomas Pospisek wrote: LXC VMs become unbootable when upgraded to jessie. The upgrade migrates systems automatically to systemd, after which the VMs won't be able to boot. Is this still relevant after the upload of lxc 1.0.6-5? Regards, Adam -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On 2014-11-23 23:19, Tomas Pospisek wrote: LXC VMs become unbootable when upgraded to jessie. The upgrade migrates systems automatically to systemd, after which the VMs won't be able to boot. The simplest workaround currently is to upgrade and then, before rebooting, to switch back to sysv. That sounds like something we should be fixing, rather than documenting. OMMV, but I'm not particularly keen on the release notes explicitly suggesting that people go out of their way not to run the default init system (quite aside from the fact that we don't currently know whether upgrades will actually change the init system at all for most users). Regards, Adam -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On Mon, 24 Nov 2014, Adam D. Barratt wrote: On 2014-11-23 23:19, Tomas Pospisek wrote: LXC VMs become unbootable when upgraded to jessie. The upgrade migrates systems automatically to systemd, after which the VMs won't be able to boot. The simplest workaround currently is to upgrade and then, before rebooting, to switch back to sysv. That sounds like something we should be fixing, rather than documenting. I agree. However nobody seems to be stepping forward to actually do the work. The question is also if that work would actually be able to migrate into jessie in time. As far as I have understood Daniel Baumann, lxc's maintainer (Cc:ed), a newer lxc version is required to make the upgrade work properly. If we leave the situation as is, then people/sysadmins will get their VM migrated to systemd and the VM will not start at its next reboot. So given the situation, my aim is to let the user know via release-notes, what he's expected to encounter and what he can do about the problem. OMMV, but I'm not particularly keen on the release notes explicitly suggesting that people go out of their way not to run the default init system. Understood and agreed. However I think it's preferable to inform the user about what is known to work instead of wishing it were different, when there's no better alternative at the moment. (quite aside from the fact that we don't currently know whether upgrades will actually change the init system at all for most users). The VM I upgraded ca. three weeks ago was switched to systemd automatically and AFAI can see, there is no component inside that VM that would depend on systemd and thus pull it in. If automatically switching the user to systemd is currently under review/discussion, could you please give me a pointer to that discussion (bug number? Thread on some mailing list?), so I can follow it and try to keep my proposed release-notes changes in sync accordingly (or drop them, in case the upgrade would not switch users to systemd by default any more when jessie is released)? Thanks greets, *t -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
On 2014-11-24 8:39, Tomas Pospisek wrote: If automatically switching the user to systemd is currently under review/discussion, could you please give me a pointer to that discussion (bug number? Thread on some mailing list?) There's tech-ctte #762194. Regards, Adam -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Bug#770776: LXC breakage and workarounds when upgrading VMs to jessie
Package: release-notes Severity: important Tags: patch Dear release-notes maintainers, LXC VMs become unbootable when upgraded to jessie. The upgrade migrates systems automatically to systemd, after which the VMs won't be able to boot. The simplest workaround currently is to upgrade and then, before rebooting, to switch back to sysv. I'm trying to collect information about this problem and possible work-arounds in the wiki [1], and will try to extract and refine a proposed text for the release-notes. My first proposed text for the release-notes is below. Please let me know if you prefer me to submit a proper patch against a SVN checkout of ddp. In chapter 4.5.7. Special care for specific packages I propose to introduce a new subchapter 4.5.7.2. LXC with the following text: The upgrade from wheezy to jessie will switch the init system to systemd. Currently LXC VMs running systemd inside are not able to boot. The workaround is to proceed with the upgrade, but *before* a reboot of the VM, switch back to the original init system. You can determine your current init system with the following command: # dpkg -S /sbin/init | sed 's/:.*//' If you are running with the default sysvinit init system, then switch back to your original init system with: # apt-get install sysvinit-core systemd-shim This will reinstall the sysvinit init system and deinstall systemd. You can find further information on LXC in Debian [[https://wiki.debian.org/LXC|in the Debian wiki]]. Thanks, *t [1] https://wiki.debian.org/LXC -- System Information: Debian Release: 7.7 APT prefers stable APT policy: (500, 'stable'), (500, 'oldstable') Architecture: amd64 (x86_64) Foreign Architectures: i386 Kernel: Linux 3.16-0.bpo.2-amd64 (SMP w/8 CPU cores) Locale: LANG=de_CH.UTF-8, LC_CTYPE=de_CH.UTF-8 (charmap=UTF-8) Shell: /bin/sh linked to /bin/dash -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
