Ganeti 2.5.1
Debian Squeeze
To reproduce the problem:
1. set up a cluster with at least 2 nodes where one of them have a volume group
smaller than the other (about half size)
2. create some instances with one or more DRBD disks making them them big
enough that the sum of their disks is bigger than 2 time the space on the
smaller node.
3. run the hbal command and balance the cluster as it says
I expect that I can still be able to make backups instead the balancer doesn't
reserve in the smaller node enough free space to generate the required
snapshots needed to make the backup of the biggest instance.
My wish is to have a couple of modifications:
1- add a cluster-wide new parameter that allows so specify the percentage of
space that have to be occupied by the snapshots. Namely, if I set it to 20, I
would like that the command "gnt-backup" generates a snapshot (for each disk of
the instance) that would be the 20% of the size of the disk of the instance.
(Currently it allocates the same size of the disk, doubling the needed free
space.. and there's no way to change this)
2- add a command line flag to hbal that if set does the following work:
a- calculates the disk size of all the instances and takes the biggest one. (comprensive of all its disks)
b- calculates the percentage of size that has to be reserved in case of a backup of that instance
c- balances the cluster making sure that in every node's volume-group there is at least the free space needed to create a snapshot of the biggest instance.
d- fires a warning if is not possible to balance the cluster and keep the amount free space required for the backup of the biggest instance (this space should be available on all the nodes).
I hope to have been clear.
Original issue reported on code.google.com by marcoli...@gmail.com on 16 Aug 2012 at 2:11
Original issue reported on code.google.com by
marcoli...@gmail.com
on 16 Aug 2012 at 2:11