[Users] simfs 2tb limit?

[Benoit Branciard]

Garith Dugmore a écrit :
> Benoit Branciard wrote:
>> option in $CTID.conf file...), but you certainly could overcome any 
>> size limit on the container root by setting some secondary mount 
>> points below the container tree, either bind mounts or plain device 
>> mounts. The /etc/vz/conf/$CTID.mount file is meant for this purpose.
>>
> I considered this but unfortunately I have a partition layout as follows:
> 
> df -Th on HN:
> Filesystem    Type    Size  Used Avail Use% Mounted on
> /dev/sda1     ext3     92G  1.4G   86G   2% /
> tmpfs        tmpfs    4.0G     0  4.0G   0% /lib/init/rw
> udev         tmpfs     10M   40K   10M   1% /dev
> tmpfs        tmpfs    4.0G     0  4.0G   0% /dev/shm
> /dev/sda3      xfs    4.7T  269G  4.5T   6% /data
> 
> ls -la /vz
> lrwxrwxrwx 1 root root 8 2008-07-07 14:30 /vz -> /data/vz
> 
> My plan then was to keep all containters in /data essentially. If I 
> understand mount correctly bind will make the whole /data available to 
> the relevant container instead of it being controlled by a specific quota.
> 
> Another option I had was to repartition and make a specific /dev/sdaX 
> available to the one container. This would work but does limit the 
> options I have - If I needed more disk space on another container at a 
> later stage I'm constrained by this setup.

If possible without critical data loss or downtime, repartitioning would 
be an option to consider. You may for example allocate some space (say 
70GB) on sda for the container's roots (mounted on /vz), and keep the 
huge /data to provide data space to containers by means of bind mounts 
on various points of their diretory tree.

But even without repartitioning, bind mounts offer sufficient 
flexibility to meet your needs. Let's say you create a directory 
/data/$VEID-bigdata on sda3. You could setup your container's 
/etc/vz/conf/$VEID.mount as following:

#!/bin/bash
[ -n "$VE_CONFFILE" -a -f "$VE_CONFFILE" ] || exit 1
. $VE_CONFFILE
mount -n --bind "/data/$VEID-bigdata" "$VE_ROOT/mybigdata" || exit 1

which grants $VEID access to the full size of sda3 under /mybigdata 
tree, without being limited by simfs (root) max capacity.

If I'm wrong, please let me know, I'm still not that familiar with 
openVZ quotas. At least this technique works well if container root 
(simfs-mounted) is on a different physical partition than that where the 
bind-mounted directory resides.


-- 
Ce message a ete verifie par MailScanner
pour des virus ou des polluriels et rien de
suspect n'a ete trouve.