This bug was fixed in the package lxd - 2.0.10-0ubuntu1~16.04.2
---
lxd (2.0.10-0ubuntu1~16.04.2) xenial; urgency=medium
* Fix regression in image update logic (LP: #1712455):
- 0005-Fix-regression-in-image-auto-update-logic.patch
-
I confirmed that the sysctl.d file is now present and functional.
** Tags removed: verification-needed verification-needed-xenial
** Tags added: verification-done verification-done-xenial
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to
Hello Christian, or anyone else affected,
Accepted lxd into xenial-proposed. The package will build now and be
available at
https://launchpad.net/ubuntu/+source/lxd/2.0.10-0ubuntu1~16.04.2 in a
few hours, and then in the -proposed repository.
Please help us by testing this new package. See
** Also affects: lxd (Ubuntu Xenial)
Importance: Undecided
Status: New
** Description changed:
- Reported by Uros Jovanovic here: https://bugs.launchpad.net/juju-
- core/+bug/1593828/comments/18
+ == SRU
+ === Rationale
+ LXD containers using systemd will use a very large amount of
** No longer affects: juju
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many open files"
To manage
** Changed in: juju
Milestone: 2.2-beta4 => 2.2-rc1
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
** Changed in: juju
Milestone: 2.2-beta3 => 2.2-beta4
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
This bug was fixed in the package lxd - 2.12-0ubuntu2
---
lxd (2.12-0ubuntu2) zesty; urgency=medium
* Increate default inotify limits to 1024 instances per uid.
(LP: #1602192)
-- Stéphane Graber Mon, 03 Apr 2017 18:51:34
-0400
** Changed in: lxd
** Changed in: juju
Milestone: 2.2-beta2 => 2.2-beta3
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
** Changed in: juju
Milestone: 2.2-beta1 => 2.2-beta2
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
>From internal discussion, for Juju, in clouds, there is no guarantee
that the images can support particular kernel parameters. For us, it is
best not to try and guess and better surface the error in the status.
** Changed in: juju
Importance: Critical => High
** Changed in: juju
** Changed in: juju
Milestone: 2.1.0 => 2.1-rc1
** Changed in: juju
Assignee: Richard Harding (rharding) => (unassigned)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when
** Changed in: juju
Status: Confirmed => Triaged
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
billy-olsen: You're right - because we don't install juju from the
package on the deployed machines the config to enable more lxd
containers doesn't get done there. We should apply it in that case.
** Changed in: juju
Status: Fix Committed => Confirmed
--
You received this bug
This bug is fixed for the local LXD provider scenario for 2.0 on Xenial
with this commit - http://bazaar.launchpad.net/~juju-
qa/ubuntu/xenial/juju/2.0.0/revision/214 - in which the juju-2.0.conf
file has the settings identified and dropped into /usr/lib/sysctl.d.
Unfortunately, this only
Can we see the commit related to the bug? Also, is there a plan to backport it
to 2.0 series as I'm not sure when we get 2.1 GA?
https://github.com/juju/juju/wiki/Juju-Release-Schedule
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
Tested w Juju 2.1 beta3 and verified I can get over 20 containers and
while it's slow does not get stuck in pending. The changes to the
profiles on the server that Michael has done appear to be working.
** Changed in: juju
Status: In Progress => Fix Committed
--
You received this bug
Not really. The problem is that bumping those sysctls to the value in
that document will work on most normal systems but will fail
dramatically on very low memory systems like some cloud instances and
ARM systems.
As for having LXD know when things will fail, it simply has not idea.
All the
Would it be reasonable to apply the settings that
https://github.com/lxc/lxd/blob/master/doc/production-setup.md suggests
when we add the first LXD to a machine?
What about interrogating the kernel about available resources and
refusing to add a container when it won't work? I can try adding 20
$ juju bootstrap foo lxd
Creating Juju controller "foo" on lxd/localhost
Looking for packaged Juju agent version 2.0.0 for amd64
No packaged binary found, preparing local Juju agent binary
To configure your system to better support LXD containers, please see:
I have filed bug #1631914 against lxd for better surfacing of the "too
many files open" error.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they
As a note, increasing fs.max-files had no effect for me.
Using the following settings I was able to deploy about 23 containers
with juju, which is higher than I could previously:
fs.inotify.max_user_watches = 524288
fs.inotify.max_user_instances = 256
These are the proposed settings to
FYI a new version of the inotify patchset was sent today for review:
https://lists.linuxfoundation.org/pipermail/containers/2016-October/037480.html
This approach is the real fix for the inotify part of this problem.
For open files, we've had a few reports of systemd occasionally
misbehaving and
** Changed in: juju
Status: Triaged => In Progress
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many
with
fs.inotify.max_user_instances = 2048
it didn't fail until the 187th container.
It did fail with a new message that I haven't seen:
[3831] x-187:error: Error calling 'lxd forkstart x-187 /var/lib/lxd/containers
/var/log/lxd/x-187/lxc.conf': err='exit status 1'
lxc 20161007100331.798
I added a "cat /proc/meminfo | grep Slab" to go-lxc-run.sh and found
this:
$ sysctl fs.inotify
fs.inotify.max_queued_events = 65536
fs.inotify.max_user_instances = 1024
fs.inotify.max_user_watches = 524288
$ ulimit -a
...
open files (-n) 1048576
...
$ go-lxc-run.sh
[0]
After rebooting with the new values, I did manage to get to launch a lot
more containers, failing at only the 232nd one.
I wonder if the issue is not the User number of open files, but the Root
number of open files, which requires a reboot to get updated.
I'll try to play around more to really
Interestingly, my baseline Kernel memory with no containers (and not
much other software) was about the same (~400MB). I'm not entirely sure
why it grew faster with the new settings, but didn't effect the
baseline.
--
You received this bug notification because you are a member of Ubuntu
Bugs,
I did try setting all of the items that are mentioned in production-
setup.md. To start with, a few of them are not reasonable.
max_user_instances defaults to 128, and we were able to see a difference
at 256, but not at 1024. Setting it to 1M seems silly.
I'll also note that my Kernel memory
Created bug #1631038 for the addition of an /etc/sysctl.d/10-juju.conf
file upping max_user_watches and max_user_instances which will increase
the number of containers we can run out of the box, but not by enough.
--
You received this bug notification because you are a member of Ubuntu
Bugs,
Note that there should be support for /etc/system/limits.d/10-juju.conf
I'm testing it now, but it may be that we can drop something in there as
well. I'll test it a bit, but if we have some tasteful defaults, maybe
we can make it work.
I think we can change from their default so instead of "*
Chatting with tych0 we were linked to:
https://github.com/lxc/lxd/blob/master/doc/production-setup.md
That notes a bunch of tweaks to run the system at scale. We need to
evaluate what items make sense to enable ootb, and what things are solid
for a prod lxd system but not a great idea for all
Some settings to tweak from tych0 on IRC:
https://github.com/lxc/lxd/blob/master/doc/production-setup.md
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD
Note from jam after some experimentation:
So having played with it for a bit, I'm more comfortable with an
/etc/sysctl.d/10-juju.conf that sets max_user_watches=512k and
max_user_instances=256 but if we want to get to 50 instances we need to
dig harder.
I can just barely get to 10 instances of
With Juju in the loop, I run into whatever limit a bit faster. I was
successful at doing:
juju bootstrap test-lxd lxd
juju deploy ubuntu
juju add-unit -n 5 ubuntu # wait for status to say everything is running
juju add-unit -n 5 ubuntu # wait for status to be happy
but then after doing one
5) Another data point, with
$ sysctl fs.inotify
fs.inotify.max_queued_events = 131072
fs.inotify.max_user_instances = 1024
fs.inotify.max_user_watches = 524288
(so max_queued_events 8x greater, and max_user_instances well above
previously established useful level), I still only get 19 containers.
Michael and I played around with some different settings, and here are
my notes.
1) Package kde-runtime seems to install
/etc/sysctl.d/30-baloo-inotify-limits.conf which sets max_user_watches to
512*1024
'slabtop' says that my baseline kernel memory is 380-420MB with no containers
We can include this in cloud-init for machines that will host
containers. However, for the lxd provider this would need to be set on
the machine running juju (the host machine). Should this be done at juju
install time?
--
You received this bug notification because you are a member of Ubuntu
** Changed in: juju
Milestone: 2.0.0 => 2.1.0
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many open
Per email thread, Juju needs to ship with expanded details:
> Stephane, I share your concerns around selecting the right knobs and
> testing accordingly but my main concern now is that I feel many people will
> hit this limitation when they deploy bigger charms in containers. Is there
> any way
** Changed in: juju-core
Assignee: Christian Muirhead (2-xtian) => (unassigned)
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start
@stgrabber, what's the status on this?
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1602192
Title:
when starting many LXD containers, they start failing to boot with
"Too many open files"
To
So there is a patchset that's been sent for review a couple of times
upstream which would tie a bunch of the inotify limits to a user
namespace. This would almost certainly fix this issue.
Until then, we can workaround it by bumping the default values for those knobs.
This bump could be done in
It sounds like the "fix" for this lies outside of LXD. Can we add the
affected packages and involve the right upstreams for this? Are we going
to pursue this purely upstream with systemd / kernel or are we going to
attempt a distro-patch of some sort?
--
You received this bug notification
It does look like max_user_instances is global and not namespaced. So
unless there's a resource leak somewhere that needs to be fixed, bumping
that limit may be the only option.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
I don't know that much about /proc/sys/fs/inotify/max_user_instances,
but apparently this isn't properly namespaced, so it is a global limit
for all containers on the host? As a workaround, lxd could ship a
/usr/lib/sysctl.d/ snippet to bump it, but I'm not sure if that has
other downsides
Apparently the relevant limit is
/proc/sys/fs/inotify/max_user_instances. This is "128" by default. When
increasing it with
sudo sysctl fs.inotify.max_user_instances=256
then the failed container reboots fine (and udev starts).
--
You received this bug notification because you are a member
47 matches
Mail list logo