Hi All,
I am trying to understand if glusterfs is still supported through gfapi.
Used as a reference this old post: https://opennebula.io/blog/experiences/native-glusterfs-image-access-for-kvm-drivers/ but when setting DISK_TYPE=GLUSTER I get invalid disk type for system datastore.
I am running: Opennebula 6.8.0.
Is it possible to instruct Opennebula to use libgfapi when using such datastores?
Thank you,
Alex
There is a storage deployment model built around the requirement of having shared filesystem mounts on the frontend and node. It does not matter the technology used for this purpose, gluster, nfs, cephfs, etc. as long as the mounts exists on the datastore paths in the frontend and virtualization nodes.
Hi thanks for the reply. I am aware that I can mount at the hosts whatever can be mounted and trick the system datastores to use my mountpoints through symlinks. I was able to configure glusterfs volumes and define a datastore to use the glusterfs mountpoint that I had configured through fstab and symlinks but this means that the libvirt is using FUSE to access the VM disk images instead the high performant Libgfapi (this can be seen from the VM xml). And this is due to libvirt not being informed that the mountpoint is a glusterfs mount. I would expect Opennebula to somehow inform libvirt about the type of storage so as to use the appropriate optimizations which are supported from qemu/libvirt. Is there any way to instruct libvirt to use libgfapi for such glusterfs datastores?
You could try and use the RAW section of the VM template to add custom information to the libvirt XML.
Thanks. Will try that and will come back if I have any luck.