[one-users] Full system datastore

[Jaime Melis]

Hi,

Can you run these commands and send us the output:


(in OpenNebula frontend)
$ for i in /var/lib/one/datastores/{,*}; do echo $i; df -h $i; done

(in a KVM hypervisor)
$ for i in /var/lib/one/datastores/{,*}; do echo $i; df -h $i; done

Also, what OpenNebula version do you have currently installed? (Distro and
package version)

Answering your questions:

> how can system datastore move from frontend to kvm? Can it be moved
manually?
By default OpenNebula assumes that the default system datastore:
/var/lib/one/datastores/0 is shared accross all the nodes using a
distributed file system. It can be set up as local to the hypervisors by
changing the TM_MAD of the system datastore to 'ssh' instead of 'shared'.

> what is the system datastore used for?
To hold volatile disks (swap disks etc created on the fly not associated
for an Image), to hold vm checkpoints for migrations/suspends and primarily
stores cloned disks.

> wouldn't it be better if system datastore be placed on ceph and is it
possible that it can be placed on ceph?
It can be placed on CephFS, yes. As long as you setup a shared filesystem
accross  all the nodes, including frontend of /var/lib/one/datastores/0 it
doesn't really matter what technology you're using: NFS, CephFS, GlusterFS,
etc...

cheers,
Jaime



On Wed, Aug 13, 2014 at 8:25 AM, <pawel.orzechowski at budikom.net> wrote:

>  Hi,
>
> Again, does anybody has some ideas about this. We need to recreate all
> VM's and it is not solution for production environment (good that it was
> just a begining of migration from XenServer) .
>
> Here are the steps we did:
>
>    - no machine can be run because ON says that system data store is full
>    (it shows it has 500GB instead of 7GB),
>    - we stopped all the VM's and then restarted frontend - it did not
>    helped
>    - the we restarted kvm - it did not helped
>    - we deleted this 500GB /var/tmp/8d187572e1d5ec159420701f9177150c file
>    - again machines were unable to start
>    - so we deleted VMs and created them again - good they had a template,
>    but unfortunately now the VM's got new IP addresses so it blocked access to
>    them from outside world due to firewall rules
>    - now after above steps system datastore has magically moved to
>    frontend and has again correct capacity of 7GB
>
> So the questions now:
>
>    - how can system datastore move from frontend to kvm? Can it be moved
>    manually?
>    - what is the system datastore used for?
>    - wouldn't it be better if system datastore be placed on ceph and is
>    it possible that it can be placed on ceph?
>
> Regards
>
> Pawel
>
> ---
>
> Paweł Orzechowskipawel.orzechowski at budikom.net*budikom.net <http://budikom.net>*
> ul. Trzy Lipy 3, GPNT, bud. C
> 80-172 Gdańsk
> tel.: +48 58 58 58 708
> email: biuro at budikom.net
>
>  W dniu 2014-08-05 08:32, Mateusz Skała napisał(a):
>
> Hi,
> I run some VM's on OpenNebula, but after few machines system datastore
> are full. As Image backend we are using ceph datastore (and there is
> enough disk space), but system datastore is probably from kvm host, from
> file /var/tmp/8d187572e1d5ec159420701f9177150c.
> How can I make some free space? It is possible to move system datastore
> on Ceph?
>
>
> _______________________________________________
> Users mailing list
> Users at lists.opennebula.org
> http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
>
>


-- 
Jaime Melis
Project Engineer
OpenNebula - Flexible Enterprise Cloud Made Simple
www.OpenNebula.org | jmelis at opennebula.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20140813/2fbbebc6/attachment-0001.htm>