OpenVZ Forum: Support » How to force a frozen VE to stop?

Home » General » Support » How to force a frozen VE to stop?

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

How to force a frozen VE to stop? [message #27976]

Tue, 04 March 2008 19:57

mcarreira
Messages: 30
Registered: February 2007

Member

Is there another way of stopping a VE not using "vzctl stop VE"?
I'm asking this, because somethig happened in my mail server that crashed it. I couldn't ping the server, and "vzctl enter VE" was not responding. It was frozen...
So, I tried to stop it with "vzctl stop VE" but, after a while, he answered "time out".
I had no other way of stopping it, so I tried to shutdown the hardware node, but he was not able to stop the mail VE, so, I had to pull down the button... Sad

This issue happened twice and I'm hoping it wouldn't happen again.
But if it would happen, what can I do to avoid shuting down everything?
Is there a way to force a frozen VE to shutdown?

Note: I'm using vzctl-3.0.16-1, kernel 2.6.18-ovz28stab035.1-smp and Centos-4.5-x86_64

Report message to a moderator

Re: How to force a frozen VE to stop? [message #27978 is a reply to message #27976]

Tue, 04 March 2008 23:17

Valmont
Messages: 225
Registered: September 2005

Senior Member

Yeah, except advice to upgrade kernel/vzctl in spec. this case. It's supposed. I really want to know too.

I have the same issue some times (very rarely) with a last kernels. In the last time unfsd (userspace daemon) in vps going crazy (high cpu uasge, but not in D state) and I could not kill it, could not stop vps.

But, I think there is no such solution, like a kill whole container's group processes, if something goes very wrong. Only reboot/reset node, yes?

Report message to a moderator

Re: How to force a frozen VE to stop? [message #28217 is a reply to message #27976]

Tue, 11 March 2008 19:59

dowdle
Messages: 261
Registered: December 2005
Location: Bozeman, Montana

Senior Member

One way to force the container to stop is to identify all of the processes in the container... and then kill them from the host node.

To find the init processes... do a "pstree -nup | grep init" and then do a vzpid on each init process until you identify the init process for the container in question.

Then do a pstree -nup and find all of the child process pids that belong to the init for the container. Killing all of the processes that fall under the init and then the init will stop the container. You might still need to umount /vz/root/{CT}.

I really don't recommend you do it that way but it can be done if needed.

--
TYL, Scott Dowdle
Belgrade, Montana, USA

Report message to a moderator