[heketi-devel] Am I missing something

[Michael Bengtson]

Hello Heketi Team,



We love using Heketi.  I have been searching far and wide for something and I was hoping I am just simply missing it, right in front of my face.  Is there a way to use Heketi to create a distributed Gluster volume—but to make sure Gluster spreads it across more than one storage peer?  I need a distributed volume (no replication) where I know the [multiple] bricks will be on different hosts.  Example: If I ask for 6G of storage, it would be nice to make the 6G logical Gluster volume spread across 3 x 2G volumes under the hood, on 3 different storage peers.  I am looking for throughput, and with my heavy writers, the single brick that got allocated to all 6G is being hit very hard, and I'd like to spread that out to <n> peers (e.g. 6G/<n>) using the Heketi api.  (Or, is this just something that I need to do manually from gluster-cli only?  When I saw 'gluster volume add-brick ...'  I was hoping that Heketi exposed that behavior somehow, either on initial allocation, or after the fact.)



Using "type: none" (distributed) I get my volume no problem, it's just that it always seems to fall to one big chunk on one single host.  I even wondered if there was a way to use the topology to hint to Heketi that it should split the bricks up between more than one host.



I have used heketi in Kubernetes by making StorageClass, and also have some experience using the heketi-cli.



Thank you for any help or pointers you can provide!



Sincerely,

Michael
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/heketi-devel/attachments/20200413/a4385578/attachment.html>