[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] cleaning up vm resources on vm shutdown when the original 'xl create' is gone



Mark Pryor <tlviewer@xxxxxxxxx> wrote [01.06.17]:
> Jim,
> Check to see if you have the default settings in/etc/default/xendomains
> then change 2 lines:
> # append these
> XENDOMAINS_SAVE=
> XENDOMAINS_RESTORE=false

I already have those.

> Is your build from git? Did you enable systemd? Its my impression thatif you 
> use the distro *deb set for xen, only the sysv init scripts are used.
> You can see a robust way to configure upstream-xen 
> here:http://107.185.103.5/xen/ubuntu/xenial/46x/source/debian/
> open the rules and see where ./configure is run. In that tree there isa 
> complete debianised build of xen-4.6 for Xenial.

I'm using xen 4.6 that comes with ubuntu 16.04.  It does use sysv init 
scripts.  The problem is systemd is killing the forked xl processes on 
hypervisor shutdown before xendomains init script has had a chance to 
run.  When xendomains script does run the best it can do is get the vm's 
into an 's' state waiting for something to clean up their resources.  

Someone suggested to me off list to xl destroy them.  For now, I'm 
just going to modify xendomain to issue an xl destroy once it sees a 
vm switch to the 's' state after its issued the xl shutdown.  That
should deal with cleaning up all the vm resources on hypervisor 
shutdown/reboot.

thanks
jim


> 
> PryMar56
>  
> 
>     On Thursday, January 5, 2017 4:52 PM, Jim Westfall 
> <jwestfall@xxxxxxxxxxxxxxxx> wrote:
>  
> 
>  Hi
> 
> I'm using xen 4.6 (ubuntu 16.04).  Looking for some guidance with the 
> following issue.
> 
> When you create a vm via 'xl create <cfg>', the xl process gets forked 
> off in to the background.  This process seems to be the one on the hook 
> for cleaning up vm resources (vif/vbd/etc) when the vm is shutdown.
> 
> However how to do you go about cleaning up the vm resources if this xl 
> process goes away (killed)?
> 
> Why they are getting killed is fun too.  When you have vm's running and 
> you shutdown/reboot the hypervisor, xendomains script is suppose to shut 
> down any active vm's.  This does cause the vm's to shutdown but the 
> vm resources are never released because systemd has already killed all 
> the forked xl processes before hand as they aren't associated with any 
> systemd service.
> 
> This causes the vm's to get stuck in an 's' state.
> 
> thanks
> jim
> 
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxx
> https://lists.xen.org/xen-users
> 
>    
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxx
> https://lists.xen.org/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
https://lists.xen.org/xen-users

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.