ITG Unix Support
>    
     |  List directory  |  History  |  Similar  |  Print version  

HPC > Cloudcity Cluster > Powering on the cloudcity cluster nodes

Powering on the cloudcity cluster nodes

When the Titan64 Nest loses power, each of the nodes are disabled. This means that the power button on the front does not actually do anything. To re-enable the nodes, you must have Angstrom Power Management Utilities installed on a machine, and connect to the PMU device. Please note that these utilities are installed on isildur and strongmad.

Starting the Power Management Daemon

We first have to start the power management daemon. We will run it on port 12001. We need to point it to the IP of the PMU, because it resolves DNS names incorrectly. Please note that this IP is different than the bryce example.

% p_control_d 10.79.133.5 12001

Take careful note of the "inet address" that it is listening on. This will be one of the IP's available on your machine, though normally you only have one. We will refer to this IP as managerip.

Powering on a node

The following example powers on nodes 1 and 2

p_control -p 12001 -h managerip -sl 1 -st 1
p_control -p 12001 -h managerip -sl 2 -st 1

Powering off a node

The following example powers off nodes 1 and 2

p_control -p 12001 -h managerip -sl 1 -st 0
p_control -p 12001 -h managerip -sl 2 -st 0

Powering on the entire nest

I recommend powering each node on one at a time, with a few seconds in between each poweron command. Start with node 1 and work your way up to node 4. For the sake of completeness, it is possible to power on the entire nest simultaneously, but I do not recommend it due to the power surge:

p_control -p 12001 -h managerip -nest -st 1

 

Reference http://wiki.chem.indiana.edu/HPC/PoweringOnTheCloudcityClusterNodes
Rights rw-rw-r--   tstrombe   ITG

Prev. About cloudcity   cloudcity power management unit settings Next