|
Hi all,
I am setting up a new gluster install to service a backup infrastructure. I am starting off with 2 nodes, each has 2 SSDs for OS and 12 x 8TB disks for data. I was planning to go RAID6 on the data vol and then breaking up the raid set into multiple LUNs
via the array software in order to present multiple luns to the OS. Then use each lun as a brick and put multiple volumes inside each brick.
The backup software I use wants to have multiple volumes of 5 – 10TB max, and many of them, rather than fewer large volumes. I will end up with 100+ volumes.
Reading many things over the last few weeks seem to suggest having many bricks per server is not a good idea especially if you will have many volumes due to the multiplication of the processes and memory especially when it needs to do healing.
What I am now planning to do is splice the RAID6 into 2 logical drives (+-35TB each) and use that as a brick, so 2 bricks per server. Then create multiple volumes on each brick and set a quota on each volume to make sure I can keep free space under control.
Then when I go from 2 nodes to 4 nodes, I will add 2 more bricks per volume, the free space on the brick will increase which will enable more volumes to be created.
Is this a good idea ? I know it will work technically, but that doesn’t mean it is the best way to do it. I don’t foresee myself using the snapshot feature so staying away from LVM at the moment if it can be helped to keep things simple.
Please, any other suggestions welcome as obviously it isn’t easy to change this type of thing once you get started.
Many Thanks
Heinrich Fourie
|
_______________________________________________ Gluster-users mailing list Gluster-users@xxxxxxxxxxx http://www.gluster.org/mailman/listinfo/gluster-users
