We're still waiting/trying to integrate the shared storage patches from
Apollon Oikonomopoulos, and once that is done, then integrating other
external storage providers will be much easier.
An alternative would be to provide a "naked" block-device support (no
management at all), that would allow use of any block device with
consistent naming across machines.
regards,
iustin
I think that this is achievable via the file-storage today plus NFS
mounting of /srv/ganeti/file-storage on nodes.
Yes, we know that the file-storage backend is not well working (free
disk space is not well compute, etc.), but confirmation whether or not
this is a good model for you would be useful.
thanks,
iustin
That's good news. Do you have any guess on when this would be
complete? If there is anything I can do to assist then I'm happy to do
so. Do you have any suggestions where be a good place to start
familiarising myself with Ganeti's architecture?
Regards,
Damien
Not sure unfortunately :(
> Do you have any suggestions where be a good place to start
> familiarising myself with Ganeti's architecture?
Reading or skimming the design documents shipped with the source tree
would be a good start, I think.
regards,
iustin
Okay thanks, I will give them a read and hopefully be in a better
position to assist at some point in the future.
With the shared storage patches in place, will file-storage work with
failover and migrate? Presently, these are restricted to drbd disk
layouts.
Thanks,
sf
Yes, they also cover file-storage (IIRC), so these three disk templates
will allow failover/migrate:
- drbd (current)
- shared block devices (SAN, Ceph in the future)
- shared filesystems (NAS/NFS/etc.)
regards,
iustin