[pkg-lxc-devel] Bug#824779: Bug#824779: container getty-static.service causes lxcfs high cpu usage

[Michael Biebl]

Am 18.07.2016 um 09:05 schrieb Evgeni Golov:
> On Mon, Jul 18, 2016 at 07:56:40AM +0200, Martin Pitt wrote:
>> Hello Michael, Evgeni,
>>
>> Evgeni Golov [2016-07-18  7:25 +0200]:
>>>>> There is no point in even wasting four getty processes on tty1-4 in
>>>>> LXC -- containers are not meant to have gettys on ttys in the first
>>>>> place. I committed a fix to git for that.
>>>>> (ConditionVirtualization=!container)
>>>
>>> this will break lxc-console, but more later :)
>>
>> console-getty.service suffices for that -- this is running in
>> containers. Interestingly I also have container-getty@{0,..3}.service
>> running. "lxc-start -n mycontainer -F" also works, i. e. you see the
>> boot messages.
> 
> Cool, will test that. Does this also apply to Jessie inside the
> container? Or only Stretch+?
> 
>>>> I just created a test jessie lxc container (using lxc 2.0.0-3~bpo8+1) on
>>>> jessie host system (using lxc-create -t download)
>>>
>>> -t debian please, -t download will download an image from
>>> linuxcontainers.org, which might or might not be the same as -t debian.
>>> (AFAIK it is almost identical, but please don't file bugs if it isn't
>>> ;))
>>
>> The "download" images are built from the exact same templates. Using
>> the "debian" template will do the exact same things, check
>> /usr/share/lxc/templates/lxc-debian . This has those modifications.

Fwiw, only the download option seems to work if you try to create
unprivileged containers.

> That was meant more as a disclaimer. :)
> The images are built from master and contain minimal changes
> (https://github.com/lxc/lxc-ci/blob/master/templates/debian.json), but
> yeah, usually the work/don't work matrix will be identical.
> 
>> I do remember that Stèphane Graber (LXC upstream) was bugging me about
>> fixing sigpwr.target in systemd proper
>> (https://lists.linuxcontainers.org/pipermail/lxc-users/2015-May/009279.html
>> and https://launchpad.net/bugs/1457321), the udev stuff isn't
>> necessary at all as udev will already not start in LXC due to
>> ConditionPathIsReadWrite=/sys, and the getty modifications should now
>> be obsolete too.
> 
> ACK @ udev
> systemd sigpwr is fixed since Stretch, so we'll have to keep that until
> the fix was backported?
> getty I'll test as mentioned above.
> 
> happy to drop as much as possible :)
> 
> Greets
> 
>>
>> The lxc-ubuntu template does none of these modifications, and it works
>> fine. So indeed they should be dropped from the lxc-debian template as
>> well. (This should be done upstream).
>>

Evgeni, I've cherry-picked some of the changes for our jessie branch:

https://anonscm.debian.org/cgit/pkg-systemd/systemd.git/log/?h=jessie
(the three most recent commits)

Can you please review those changes and let me know if there is
something missing?

Regards,
Michael

-- 
Why is it that all of the instruments seeking intelligent life in the
universe are pointed away from Earth?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-lxc-devel/attachments/20160718/92be6c20/attachment.sig>