2

I'm trying to configure a proof of concept where a server automatically gets spun up and joins a replication cluster. When the server is no longer needed, it gets spun down, and automatically removed from the cluster.

There is no way of pre-defining IP addresses, nor is there a way of controlling which servers get spun down when less servers are needed. (Amazon Auto-Scaling)

How can I get the server to discover and join the already existent Gluster cluster? I was thinking of running a cron job @reboot, which probes the network to see if there are any servers online running Gluster, and then SSH's into the first it finds, executing the necessary commands. The problem is that I can't seem to find a way to succesfully probe the network.

To make automated shutdowns a not an issue, I was thinking of running a cron job on each of the cluster peers, checking if there are any disconnected peers, and removing those disconnected peers if any are found. This seems quite convoluted. Is there a better/cleaner way of doing this?

I'm rather new to GlusterFS. Any tips and pointers are welcome.

Simon
  • 193
  • 2
  • 10

2 Answers2

0

After searching a while, I discovered 'AWS CLI' in order to find a set of instances with a given tag, or in a given scaling group. From there on, it's possible to pick one, SSH in, and do any necessary commands.

Simon
  • 193
  • 2
  • 10
0

I encountered the same issues while trying to make GlusterFS work with AWS. Disconnecting servers isn't as much of an issue as making new server joins the cluster automatically.

GlusterFS requires that we always know the IP/hostname of the "other servers" to peer, and it's a limitation if we want to autoscale.

Viet
  • 1