I have a simple GlusterFS 7.2 setup with 3 servers:
- data1 and data2 contribute several bricks (4 each) to a distributed replicated GlusterFS volume.
- host1 uses the GlusterFS FUSE client to mount the shared filesystem, but does not contribute storage space.
The setup I'm using pretty much follows the quick start guide: https://docs.gluster.org/en/latest/Quick-Start-Guide/Quickstart/#installing-glusterfs-a-quick-start-guide. I am using 2x replication (I know 3x would be better to avoid split brains).
Even when no process is writing to/reading from the shared filesystem on any of the servers, I am seeing large network usage (with nload or iftop) between data1 and data2. This may make a bit of sense considering the replication setup I am using (even though it seems weird).
However, I am also seeing heavy usage on host1 (dozens of Mbps), which is almost 100% due to a glusterfs process communicating with data1 and data2. Since I am simply mounting that disk, I don't see why there should be such usage.
Is anything wrong?
