check this doc
//github.com/prashmohan/lxc-fork/blob/master/Documentation/filesystems/nfs/nfsroot.txt
//github.com/prashmohan/lxc-fork/blob/master/Documentation/filesystems/nfs/nfsroot.txt
This is not what we need. The Image file on NFS is for sure possible and working. We like to use real nfsroot like KVM with pxeboot.
I understand dietmar that nfsroot is possible using the nfsroot kernel parameter so that / of the container is mounted via nfs.
however i am not sure if the host needs to mount all roots for its guests and then maps the path to the containers or the containers really can mount themselves so that's why i am asking... using the nfsroot=NAS:/path/to/roor parameter we can directly mount from the container context which is very nice because we do not clutter the hosts IO queues as we only have one per context ....
Is it possible that the root filesystem ist not an image file and the files are accessible like in former times with OpenVZ?
So we could simple mount a NFS volume that hosts the root filesystems?
So we could simple mount a NFS volume that hosts the root filesystems?
It is also possible with NFS if you create a root disk using size 0. This case simply creates a subdir and use it as container root.
Please note that storage size is not restricted in this case.
Please note that storage size is not restricted in this case.
NFS IO is then handled by host OS only, so no network IO to the NFS storage network is required for the container, right?
I get an error with disk size 0 - The minimum value for this field is 0.1.
So is this not possible with the gui?
So is this not possible with the gui?
Are there known problems with this configuration?
Will this option be removed in the next releases?
Will this option be removed in the next releases?
Btw, there is a "pct mount' command, which can mount the container root on the host. This also works if you use image files for containers.
We use a NetApp Metrocluster as storage system so i don't think we have a problem with the storage backend. Most requests will be answered from cache and not from disk. We have about 33% from nvram 33% from FlashCache and 33% from disk.
We have 2x 10G links to each hypervisor so this should be enough. We also use ultra low latency switches in the backend.
The max volume size is 100TB so size is also no problem.
The pct mount does not because we need the snapshots of the files every hour. If some mistake happen we simple copy the file out of the .snapshot directory. This is why we need this solution.
We have 2x 10G links to each hypervisor so this should be enough. We also use ultra low latency switches in the backend.
The max volume size is 100TB so size is also no problem.
The pct mount does not because we need the snapshots of the files every hour. If some mistake happen we simple copy the file out of the .snapshot directory. This is why we need this solution.