A new week-end project appears!

- Create 3 #NetBSD VMs, each with 3 disks, 1 system under FFS, 2 and 3 ZFS (or FFS + LVM0.

- Install gluster on all 3 VMs.

- Start storing data on the VMs, observe and measure replication and performances.

- Stop a member of the gluster cluster and see what happens.

At least #pkgsrc is pretty clear about gluster requirements: 🤔

*** GlusterFS servers ***

glusterfs servers will need extended attributes in exported filesystems.

This is currently only supported in UFS1 (newfs -O1). You will need to create the directories below on exported filesystems' roots, and then mount with -o extattr
.attribute/system
.attribute/user

So, no ZFS for this little experiment.

This also seriously limits the size of the gluster filesystem.

And, of course, I now have to upgrade qemu by compiling version 7.1.0.

Just to make sure I have something a bit better than 7.0.0 - which is already great.

In my defence, I am just doing a "configure" and not trying to fiddle with all the bits and pieces of potential configuration.

Whcih also means I am compiling ALL THE QEMU TARGETS. Should take a while...

[1668/9779] Compiling C object...

Right, this may take a while.

I love the fact that open-source provides super stable, powerful solutions for that kind of experiment.

That little machine runs super stable and cool under Slackware.