OpenVZ Forum: Support » Issue with stopping VE in Ubuntu 8.04

Home » General » Support » Issue with stopping VE in Ubuntu 8.04

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Issue with stopping VE in Ubuntu 8.04 [message #32619]

Tue, 19 August 2008 17:20

nixgeek
Messages: 5
Registered: August 2008

Junior Member

Hi,

I have been very excited about OpenVZ and its similarities with Sun Zones.
However, I am having an issue with stopping VE's after I have
created them. After it fails I am unable to enter the VE.
Nor can I reboot. I have to manually reboot it.

I am running the latest version.

Server: Ubuntu 8.04
Vzkernel: Linux vmserver 2.6.24-19-openvz #1 SMP Sat Jul 12 01:13:47 UTC 2008 x86_64 GNU/Linux

vzctl: 3.0.22
VE: ubuntu-8.04-amd64-minimal

Hardware:
CPU: Intel core 2 duo e6400
Mem: 4GB
Nic: Intel Pro 1000 ethernet adapter

I am using the cached template from:
http://wiki.openvz.org/Download/template/precreated

Attached is the conf file I am using to create the VE.
It is based on vzsplit.

Any ideas?

This is stopping my progression in this adventure.

Attachment: ve-64hosts.conf-sample
(Size: 0.98KB, Downloaded 260 times)

[Updated on: Tue, 19 August 2008 17:22]

Report message to a moderator

Re: Issue with stopping VE in Ubuntu 8.04 [message #32628 is a reply to message #32619]

Wed, 20 August 2008 22:12

piavlo
Messages: 159
Registered: January 2007

Senior Member

First of all 2.6.24 is unstable branch, you can try more stable 2.6.18

Check if you have any processes which are in D state
(uninterruptable sleep) inside of VE or kernel threads (like events) which are also in D state.
IMHO this the MAIN HUGE problem with OpenVZ - once you get one of those D state processes you can stop the VE
also you actually can't stop the HN also. This happens a lot with
processes accessing nfs shares but only inside VE, while there are no problems with remote nfs server.
Once OpenVZ releases kernel 2.6.25+
this will be partially resolved, as there are new TASK_KILLABLE
state which allows processes in this state to be killed with SIGKILL, and nfs client code is already utilizing this new state, which will allow those VE to be stopped (only if the stuck processes are using TASK_KILLABLE sleep).

Report message to a moderator

Re: Issue with stopping VE in Ubuntu 8.04 [message #32632 is a reply to message #32628]

Thu, 21 August 2008 02:36

nixgeek
Messages: 5
Registered: August 2008

Junior Member

Ah interesting...

I am using NFS in the VE.

And it does appear after I mount the NFS share that
this issue arises. However, I've tried to umount
the shares prior to shutting it down. Still same issue.
So it might be a /etc/init.d/portmap not stopping or
some other NFS related app.

I noticed that there are 2-3 process still running even
after the 'vzctl stop' command times out.

After that I am hosed... Cannot even reboot. I have to do
a 'halt -fp' in the HW node.

I did not know this was the unstable release.

Thanks for the info...

Report message to a moderator

Re: Issue with stopping VE in Ubuntu 8.04 [message #32634 is a reply to message #32628]

Thu, 21 August 2008 02:51

nixgeek
Messages: 5
Registered: August 2008

Junior Member

Thanks for the info...

So what is the best way to deal with this in the meantime?

What is the work-a-round? {if any}

Report message to a moderator

Re: Issue with stopping VE in Ubuntu 8.04 [message #32637 is a reply to message #32634]

Thu, 21 August 2008 11:25

piavlo
Messages: 159
Registered: January 2007

Senior Member

Even with so called stable 2.6.18 release there are same problems.
When a process get stuck in D state inside VE (probably trying to access files on nfs mounts), there are usually corresponding one of the [events/cpunum] kernel thread in D state, meaning the only way
is to hard reset the node (or little better use MagicSysRq for cleaner reset).
One of the things i've noticed that makes these hangup D processes
to appear much less frequently, is that you should run portmap on HN too (since i've noticed there are HN dmesg notifications smthing like portmap not responding, i don't know if it's regaring portmap not reponding inside VE of the HN itself, since there is one kernel
and one dmesg, and i can't if a specific dmessage is related to some VE or the HN itself). Another thing i did ,out of paranoia, is just install nfs-utils on HN too. I'd love to hear from the developers what they think about this portmap issue?

In general the nfs support in OpenVZ is very BAD, i've bugged the
developers several time about a year ago, but nothing changed since. Since i've got the feeling that nfs support is not important feature for them i've not bugged them since then.

Report message to a moderator

Previous Topic:	How to install guest OS on container
Next Topic:	HELP! can not find "which" command in VPS

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Mon Jul 15 08:08:44 GMT 2024

Total time taken to generate the page: 0.02340 seconds