Large memory clusters

[mflynn-lanl]

@edwins
I created a large memory cluster (snapshot from a large memory VM) and used a large memory VM type for the worker node, but I didn't include a large enough volume to store the Cromwell executions data. So I created a new one with more scratch space but that one errored during creation. I created another cluster, this time using the same VM snapshot I used for our regular cluster, and a large-memory (r3xl) for the worker node. This cluster was created successfully, I could ssh to it and call sinfo but at some point later on when I tried to ssh in again, I was not able to because the ssh login timed out

[mflynn-lanl]

This is not an issue for clusters made with m3.2xl worker nodes

[mflynn-lanl]

It is working now. I can create a cluster with a r3.xl worker node. Not sure why it didn't work before using the same deployment parameters

[edwins]

one thing to consider is that when you create a snapshot, any future vms must be the size of the vm you snapshotted or larger. I think you might be hitting upon that issue.

A good practice is if you need to create a snapshot, launch a vm with the smallest size possible. that way, you will have flexibility in launching other sizes in the future.