Frequently Asked Questions (FAQ)

The following is a set of common questions and cross-reference pointers to the answers in the Scyld ClusterWare documentation.

Software Install/Update

How do I install or update ClusterWare RPMs?

Always use scyld-install to install or update the basic ClusterWare packages. See Installation and Upgrade of Scyld ClusterWare and Updating Base Distribution Software.

For optional ClusterWare packages that are not managed by scyld-install, see Additional Software.

Use a simple yum install or yum update to install or update non-ClusterWare base distribution packages.

How do I install or update software without head node Internet access?

Cluster Management

What if all ``scyld-*`` commands fail?

One reason may be the root filesystem is full. See Head Node Filesystem Is 100% Full.

Another reason may be the etcd database exceeds its size limit. See etcd Database Exceeds Size Limit.

What are hardware requirements for Scyld ClusterWare?

How do I add a compute node?

How do I replace a compute node?

How do I configure multiple head nodes?

How do I configure a job scheduler, like Slurm, TORQUE, or OpenPBS?

How do I install and configure OpenMPI?

How do I keep the host keys consistent across all compute nodes?

How do I change a node name?

How do I change IP addresses?

Manipulating Compute Node Images

How do I create an image containing a non-default kernel?

How do I recreate the default image, boot config, and attributes?

How do I create an image containing a non-default base distribution?

How do I delete unused images or boot configurations to free storage space?

Issues with Interacting with Compute Nodes

What if all ``scyld-*`` commands fail?

One reason may be the etcd database exceeding its size limit. See etcd Database Exceeds Size Limit.

Why does ``scyld-nodectl -i <NODE_NAME> ssh`` fail?

Why does ``scyld-nodectl -i <NODE_NAME> shutdown`` or ``reboot`` fail?